Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1 | /* |
Abe Mbise | 1b99338 | 2017-12-19 13:51:59 +0000 | [diff] [blame] | 2 | * Copyright (c) 2016-2018 ARM Limited. |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 3 | * |
| 4 | * SPDX-License-Identifier: MIT |
| 5 | * |
| 6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
| 7 | * of this software and associated documentation files (the "Software"), to |
| 8 | * deal in the Software without restriction, including without limitation the |
| 9 | * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or |
| 10 | * sell copies of the Software, and to permit persons to whom the Software is |
| 11 | * furnished to do so, subject to the following conditions: |
| 12 | * |
| 13 | * The above copyright notice and this permission notice shall be included in all |
| 14 | * copies or substantial portions of the Software. |
| 15 | * |
| 16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
| 19 | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
| 22 | * SOFTWARE. |
| 23 | */ |
| 24 | #include "arm_compute/core/NEON/kernels/NECannyEdgeKernel.h" |
| 25 | |
| 26 | #include "arm_compute/core/AccessWindowStatic.h" |
| 27 | #include "arm_compute/core/Error.h" |
| 28 | #include "arm_compute/core/Helpers.h" |
| 29 | #include "arm_compute/core/ITensor.h" |
| 30 | #include "arm_compute/core/TensorInfo.h" |
| 31 | #include "arm_compute/core/Types.h" |
| 32 | #include "arm_compute/core/Utils.h" |
| 33 | #include "arm_compute/core/Validate.h" |
| 34 | |
| 35 | #include <arm_neon.h> |
| 36 | #include <cstddef> |
| 37 | #include <cstdint> |
| 38 | #include <tuple> |
| 39 | |
| 40 | using namespace arm_compute; |
| 41 | |
| 42 | namespace arm_compute |
| 43 | { |
| 44 | class Coordinates; |
| 45 | } // namespace arm_compute |
| 46 | |
| 47 | namespace |
| 48 | { |
| 49 | constexpr int NO_EDGE = 0; |
| 50 | constexpr int EDGE = 255; |
| 51 | constexpr int MAYBE = 127; |
| 52 | } // namespace |
| 53 | |
Ioan-Cristian Szabo | 5edbd1c | 2017-11-13 13:34:08 +0000 | [diff] [blame] | 54 | #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 55 | namespace fp16 |
| 56 | { |
| 57 | inline uint8x8_t phase_quantization(const float32x4x2_t &gx, const float32x4x2_t &gy) |
| 58 | { |
| 59 | // Constant use for evaluating score1 and score3 |
| 60 | static const float32x4_t const45 = vdupq_n_f32(0.70710678118655f); |
| 61 | static const float32x4_t zero = vdupq_n_f32(0.0f); |
| 62 | static const float32x4_t one = vdupq_n_f32(1.0f); |
| 63 | static const float32x4_t two = vdupq_n_f32(2.0f); |
| 64 | static const float32x4_t three = vdupq_n_f32(3.0f); |
| 65 | |
| 66 | // Score0: (1, 0) |
| 67 | const float32x4x2_t score0 = |
| 68 | { |
| 69 | vabsq_f32(gx.val[0]), |
| 70 | vabsq_f32(gx.val[1]) |
| 71 | }; |
| 72 | |
| 73 | // Score2: ( 0, 1 ) |
| 74 | const float32x4x2_t score2 = |
| 75 | { |
| 76 | vabsq_f32(gy.val[0]), |
| 77 | vabsq_f32(gy.val[1]) |
| 78 | }; |
| 79 | |
| 80 | // Score1 and Score3: ( sqrt(2) / 2, sqrt(2) / 2 ) - ( -sqrt(2) / 2, sqrt(2) / 2 ) |
| 81 | float32x4x2_t score1 = |
| 82 | { |
| 83 | vmulq_f32(gy.val[0], const45), |
| 84 | vmulq_f32(gy.val[1], const45) |
| 85 | }; |
| 86 | |
| 87 | float32x4x2_t score3 = score1; |
| 88 | |
| 89 | score1.val[0] = vmlaq_f32(score1.val[0], gx.val[0], const45); |
| 90 | score1.val[1] = vmlaq_f32(score1.val[1], gx.val[1], const45); |
| 91 | score3.val[0] = vmlsq_f32(score3.val[0], gx.val[0], const45); |
| 92 | score3.val[1] = vmlsq_f32(score3.val[1], gx.val[1], const45); |
| 93 | |
| 94 | score1.val[0] = vabsq_f32(score1.val[0]); |
| 95 | score1.val[1] = vabsq_f32(score1.val[1]); |
| 96 | score3.val[0] = vabsq_f32(score3.val[0]); |
| 97 | score3.val[1] = vabsq_f32(score3.val[1]); |
| 98 | |
| 99 | float32x4x2_t phase = |
| 100 | { |
| 101 | zero, |
| 102 | zero |
| 103 | }; |
| 104 | |
| 105 | float32x4x2_t old_score = score0; |
| 106 | |
| 107 | // score1 > old_score? |
| 108 | uint32x4x2_t mask = |
| 109 | { |
| 110 | vcgtq_f32(score1.val[0], old_score.val[0]), |
| 111 | vcgtq_f32(score1.val[1], old_score.val[1]) |
| 112 | }; |
| 113 | |
| 114 | phase.val[0] = vbslq_f32(mask.val[0], one, phase.val[0]); |
| 115 | phase.val[1] = vbslq_f32(mask.val[1], one, phase.val[1]); |
| 116 | old_score.val[0] = vbslq_f32(mask.val[0], score1.val[0], old_score.val[0]); |
| 117 | old_score.val[1] = vbslq_f32(mask.val[1], score1.val[1], old_score.val[1]); |
| 118 | |
| 119 | // score2 > old_score? |
| 120 | mask.val[0] = vcgtq_f32(score2.val[0], old_score.val[0]); |
| 121 | mask.val[1] = vcgtq_f32(score2.val[1], old_score.val[1]); |
| 122 | |
| 123 | phase.val[0] = vbslq_f32(mask.val[0], two, phase.val[0]); |
| 124 | phase.val[1] = vbslq_f32(mask.val[1], two, phase.val[1]); |
| 125 | old_score.val[0] = vbslq_f32(mask.val[0], score2.val[0], old_score.val[0]); |
| 126 | old_score.val[1] = vbslq_f32(mask.val[1], score2.val[1], old_score.val[1]); |
| 127 | |
| 128 | // score3 > old_score? |
| 129 | mask.val[0] = vcgtq_f32(score3.val[0], old_score.val[0]); |
| 130 | mask.val[1] = vcgtq_f32(score3.val[1], old_score.val[1]); |
| 131 | |
| 132 | phase.val[0] = vbslq_f32(mask.val[0], three, phase.val[0]); |
| 133 | phase.val[1] = vbslq_f32(mask.val[1], three, phase.val[1]); |
| 134 | old_score.val[0] = vbslq_f32(mask.val[0], score3.val[0], old_score.val[0]); |
| 135 | old_score.val[1] = vbslq_f32(mask.val[1], score3.val[1], old_score.val[1]); |
| 136 | |
| 137 | // Convert from float32x4_t to uint8x8_t |
| 138 | return vmovn_u16(vcombine_u16(vmovn_u32(vcvtq_u32_f32(phase.val[0])), |
| 139 | vmovn_u32(vcvtq_u32_f32(phase.val[1])))); |
| 140 | } |
| 141 | |
| 142 | inline uint8x8_t phase_quantization(float16x8_t gx, float16x8_t gy) |
| 143 | { |
| 144 | // Constant use for evaluating score1 and score3 |
| 145 | static const float16x8_t const45 = vdupq_n_f16(0.70710678118655f); |
| 146 | static const float16x8_t zero = vdupq_n_f16(0.0f); |
| 147 | static const float16x8_t one = vdupq_n_f16(1.0f); |
| 148 | static const float16x8_t two = vdupq_n_f16(2.0f); |
| 149 | static const float16x8_t three = vdupq_n_f16(3.0f); |
| 150 | |
| 151 | // Score0: (1, 0) |
| 152 | const float16x8_t score0 = vabsq_f16(gx); |
| 153 | |
| 154 | // Score2: ( 0, 1 ) |
| 155 | const float16x8_t score2 = vabsq_f16(gy); |
| 156 | |
| 157 | // Score1 and Score3: ( sqrt(2) / 2, sqrt(2) / 2 ) - ( -sqrt(2) / 2, sqrt(2) / 2 ) |
| 158 | float16x8_t score1 = vmulq_f16(gy, const45); |
| 159 | float16x8_t score3 = score1; |
| 160 | |
| 161 | score1 = vfmaq_f16(score1, gx, const45); |
| 162 | score3 = vfmsq_f16(score3, gx, const45); |
| 163 | |
| 164 | score1 = vabsq_f16(score1); |
| 165 | score3 = vabsq_f16(score3); |
| 166 | |
| 167 | float16x8_t phase = zero; |
| 168 | float16x8_t old_score = score0; |
| 169 | |
| 170 | // score1 > old_score? |
| 171 | uint16x8_t mask = vcgtq_f16(score1, old_score); |
| 172 | |
| 173 | phase = vbslq_f16(mask, one, phase); |
| 174 | old_score = vbslq_f16(mask, score1, old_score); |
| 175 | |
| 176 | // score2 > old_score? |
| 177 | mask = vcgtq_f16(score2, old_score); |
| 178 | |
| 179 | phase = vbslq_f16(mask, two, phase); |
| 180 | old_score = vbslq_f16(mask, score2, old_score); |
| 181 | |
| 182 | // score3 > old_score? |
| 183 | mask = vcgtq_f16(score3, old_score); |
| 184 | |
| 185 | phase = vbslq_f16(mask, three, phase); |
| 186 | |
| 187 | // Convert from float16x8_t to uint8x8_t |
| 188 | return vmovn_u16(vcvtq_u16_f16(phase)); |
| 189 | } |
| 190 | |
| 191 | /** Computes the gradient phase if gradient_size = 3 or 5. The output is quantized. |
| 192 | * 0 = 0°, 1 = 45°, 2 = 90°, 3 = 135° |
| 193 | * |
| 194 | * @param[in] gx Gx component |
| 195 | * @param[in] gy Gy component |
| 196 | * |
| 197 | * @return quantized phase for 8 pixels |
| 198 | */ |
| 199 | inline uint8x8_t phase_quantization_S16_S16(int16x8_t gx, int16x8_t gy) |
| 200 | { |
| 201 | return phase_quantization(vcvtq_f16_s16(gx), vcvtq_f16_s16(gy)); |
| 202 | } |
| 203 | |
| 204 | /** Computes the gradient phase if gradient_size = 7. The output is quantized. |
| 205 | * 0 = 0°, 1 = 45°, 2 = 90°, 3 = 135° |
| 206 | * |
| 207 | * @param[in] gx Gx component |
| 208 | * @param[in] gy Gy component |
| 209 | * |
| 210 | * @return quantized phase for 8 pixels |
| 211 | */ |
| 212 | inline uint8x8_t phase_quantization_S32_S32(const int32x4x2_t &gx, const int32x4x2_t &gy) |
| 213 | { |
| 214 | // Convert to float |
| 215 | const float32x4x2_t gx_f32 = |
| 216 | { |
| 217 | vcvtq_f32_s32(gx.val[0]), |
| 218 | vcvtq_f32_s32(gx.val[1]) |
| 219 | }; |
| 220 | |
| 221 | const float32x4x2_t gy_f32 = |
| 222 | { |
| 223 | vcvtq_f32_s32(gy.val[0]), |
| 224 | vcvtq_f32_s32(gy.val[1]) |
| 225 | }; |
| 226 | |
| 227 | return phase_quantization(gx_f32, gy_f32); |
| 228 | } |
| 229 | |
| 230 | /** Computes the magnitude using the L1-norm type if gradient_size = 3 or 5 |
| 231 | * |
| 232 | * @param[in] gx Gx component |
| 233 | * @param[in] gy Gy component |
| 234 | * |
| 235 | * @return magnitude for 8 pixels |
| 236 | */ |
| 237 | inline uint16x8_t mag_l1_S16_S16(int16x8_t gx, int16x8_t gy) |
| 238 | { |
| 239 | return vaddq_u16(vreinterpretq_u16_s16(vabsq_s16(gx)), |
| 240 | vreinterpretq_u16_s16(vabsq_s16(gy))); |
| 241 | } |
| 242 | |
| 243 | /** Computes the magnitude using the L1-norm type if gradient_size = 7 |
| 244 | * |
| 245 | * @param[in] gx Gx component |
| 246 | * @param[in] gy Gy component |
| 247 | * |
| 248 | * @return magnitude for 8 pixels |
| 249 | */ |
| 250 | inline uint32x4x2_t mag_l1_S32_S32(const int32x4x2_t &gx, const int32x4x2_t &gy) |
| 251 | { |
| 252 | const uint32x4x2_t gx_abs = |
| 253 | { |
| 254 | vreinterpretq_u32_s32(vabsq_s32(gx.val[0])), |
| 255 | vreinterpretq_u32_s32(vabsq_s32(gx.val[1])) |
| 256 | }; |
| 257 | |
| 258 | const uint32x4x2_t gy_abs = |
| 259 | { |
| 260 | vreinterpretq_u32_s32(vabsq_s32(gy.val[0])), |
| 261 | vreinterpretq_u32_s32(vabsq_s32(gy.val[1])) |
| 262 | }; |
| 263 | |
| 264 | const uint32x4x2_t out = |
| 265 | { |
| 266 | vaddq_u32(gx_abs.val[0], gy_abs.val[0]), |
| 267 | vaddq_u32(gx_abs.val[1], gy_abs.val[1]) |
| 268 | }; |
| 269 | |
| 270 | return out; |
| 271 | } |
| 272 | |
| 273 | inline float32x4x2_t mag_l2(const float32x4x2_t &gx, const float32x4x2_t &gy) |
| 274 | { |
| 275 | // x^2 ... |
| 276 | float32x4x2_t mag = |
| 277 | { |
| 278 | vmulq_f32(gx.val[0], gx.val[0]), |
| 279 | vmulq_f32(gx.val[1], gx.val[1]) |
| 280 | }; |
| 281 | |
| 282 | // ... + y^2 |
| 283 | mag.val[0] = vmlaq_f32(mag.val[0], gy.val[0], gy.val[0]); |
| 284 | mag.val[1] = vmlaq_f32(mag.val[1], gy.val[1], gy.val[1]); |
| 285 | |
| 286 | // sqrt(...) |
| 287 | mag.val[0] = vmulq_f32(vrsqrteq_f32(mag.val[0]), mag.val[0]); |
| 288 | mag.val[1] = vmulq_f32(vrsqrteq_f32(mag.val[1]), mag.val[1]); |
| 289 | |
| 290 | return mag; |
| 291 | } |
| 292 | |
| 293 | inline float16x8_t mag_l2(float16x8_t gx, float16x8_t gy) |
| 294 | { |
| 295 | // x^2 ... |
| 296 | float16x8_t mag = vmulq_f16(gx, gx); |
| 297 | |
| 298 | // ... + y^2 |
| 299 | mag = vfmaq_f16(mag, gy, gy); |
| 300 | |
| 301 | // sqrt(...) |
| 302 | mag = vmulq_f16(vrsqrteq_f16(mag), mag); |
| 303 | |
| 304 | return mag; |
| 305 | } |
| 306 | |
| 307 | /** Computes the magnitude using L2-norm if gradient_size = 3 or 5 |
| 308 | * |
| 309 | * @param[in] gx Gx component |
| 310 | * @param[in] gy Gy component |
| 311 | * |
| 312 | * @return magnitude for 8 pixels |
| 313 | */ |
| 314 | inline uint16x8_t mag_l2_S16_S16(int16x8_t gx, int16x8_t gy) |
| 315 | { |
| 316 | /* Compute magnitude using L2 normalization */ |
| 317 | const float16x8_t gx2 = vcvtq_f16_s16(gx); |
| 318 | const float16x8_t gy2 = vcvtq_f16_s16(gy); |
| 319 | const float16x8_t mag = mag_l2(gx2, gy2); |
| 320 | |
| 321 | /* Store magnitude - Convert to uint16x8 */ |
| 322 | return vcvtq_u16_f16(mag); |
| 323 | } |
| 324 | |
| 325 | /** Computes the magnitude using L2-norm if gradient_size = 7 |
| 326 | * |
| 327 | * @param[in] gx Gx component |
| 328 | * @param[in] gy Gy component |
| 329 | * |
| 330 | * @return magnitude for 8 pixels |
| 331 | */ |
| 332 | inline uint32x4x2_t mag_l2_S32_S32(const int32x4x2_t &gx, const int32x4x2_t &gy) |
| 333 | { |
| 334 | // Compute magnitude using L2 normalization |
| 335 | float32x4x2_t gx2 = |
| 336 | { |
| 337 | vcvtq_f32_s32(gx.val[0]), |
| 338 | vcvtq_f32_s32(gx.val[1]) |
| 339 | }; |
| 340 | |
| 341 | float32x4x2_t gy2 = |
| 342 | { |
| 343 | vcvtq_f32_s32(gy.val[0]), |
| 344 | vcvtq_f32_s32(gy.val[1]) |
| 345 | }; |
| 346 | |
| 347 | const float32x4x2_t mag = mag_l2(gx2, gy2); |
| 348 | const uint32x4x2_t mag32 = |
| 349 | { |
| 350 | vcvtq_u32_f32(mag.val[0]), |
| 351 | vcvtq_u32_f32(mag.val[1]) |
| 352 | }; |
| 353 | |
| 354 | return mag32; |
| 355 | } |
| 356 | |
| 357 | /** Gradient function used when the gradient size = 3 or 5 and when the norm_type = L1-norm |
| 358 | * |
| 359 | * @param[in] in1_ptr Pointer to source image. Gx image. Data type supported S16 |
| 360 | * @param[in] in2_ptr Pointer to source image. Gy image. Data type supported S16 |
| 361 | * @param[out] out1_ptr Pointer to destination image. Magnitude. Data type supported U16 |
| 362 | * @param[out] out2_ptr Pointer to destination image. Quantized phase. Data type supported U8 |
| 363 | */ |
| 364 | void mag_phase_l1norm_S16_S16_U16_U8(const void *__restrict in1_ptr, const void *__restrict in2_ptr, void *__restrict out1_ptr, void *__restrict out2_ptr) |
| 365 | { |
| 366 | const auto in1 = static_cast<const int16_t *__restrict>(in1_ptr); |
| 367 | const auto in2 = static_cast<const int16_t *__restrict>(in2_ptr); |
| 368 | const auto out1 = static_cast<uint16_t *__restrict>(out1_ptr); |
| 369 | const auto out2 = static_cast<uint8_t *__restrict>(out2_ptr); |
| 370 | |
| 371 | const int16x8x4_t gx = |
| 372 | { |
| 373 | vld1q_s16(in1), |
| 374 | vld1q_s16(in1 + 8), |
| 375 | vld1q_s16(in1 + 16), |
| 376 | vld1q_s16(in1 + 24) |
| 377 | }; |
| 378 | |
| 379 | const int16x8x4_t gy = |
| 380 | { |
| 381 | vld1q_s16(in2), |
| 382 | vld1q_s16(in2 + 8), |
| 383 | vld1q_s16(in2 + 16), |
| 384 | vld1q_s16(in2 + 24) |
| 385 | }; |
| 386 | |
| 387 | // Compute and store phase |
| 388 | vst1_u8(out2 + 0, phase_quantization_S16_S16(gx.val[0], gy.val[0])); |
| 389 | vst1_u8(out2 + 8, phase_quantization_S16_S16(gx.val[1], gy.val[1])); |
| 390 | vst1_u8(out2 + 16, phase_quantization_S16_S16(gx.val[2], gy.val[2])); |
| 391 | vst1_u8(out2 + 24, phase_quantization_S16_S16(gx.val[3], gy.val[3])); |
| 392 | |
| 393 | // Compute ans store magnitude using L1 normalization |
| 394 | vst1q_u16(out1 + 0, mag_l1_S16_S16(gx.val[0], gy.val[0])); |
| 395 | vst1q_u16(out1 + 8, mag_l1_S16_S16(gx.val[1], gy.val[1])); |
| 396 | vst1q_u16(out1 + 16, mag_l1_S16_S16(gx.val[2], gy.val[2])); |
| 397 | vst1q_u16(out1 + 24, mag_l1_S16_S16(gx.val[3], gy.val[3])); |
| 398 | } |
| 399 | |
| 400 | /** Gradient function used when the gradient size = 3 or 5 and when the norm_type = L2-norm |
| 401 | * |
| 402 | * @param[in] in1_ptr Pointer to source image. Gx image. Data type supported S16 |
| 403 | * @param[in] in2_ptr Pointer to source image. Gy image. Data type supported S16 |
| 404 | * @param[out] out1_ptr Pointer to destination image. Magnitude. Data type supported U16 |
| 405 | * @param[out] out2_ptr Pointer to destination image. Quantized phase. Data type supported U8 |
| 406 | */ |
| 407 | void mag_phase_l2norm_S16_S16_U16_U8(const void *__restrict in1_ptr, const void *__restrict in2_ptr, void *__restrict out1_ptr, void *__restrict out2_ptr) |
| 408 | { |
| 409 | const auto in1 = static_cast<const int16_t *__restrict>(in1_ptr); |
| 410 | const auto in2 = static_cast<const int16_t *__restrict>(in2_ptr); |
| 411 | const auto out1 = static_cast<uint16_t *__restrict>(out1_ptr); |
| 412 | const auto out2 = static_cast<uint8_t *__restrict>(out2_ptr); |
| 413 | |
| 414 | const int16x8x4_t gx = |
| 415 | { |
| 416 | vld1q_s16(in1), |
| 417 | vld1q_s16(in1 + 8), |
| 418 | vld1q_s16(in1 + 16), |
| 419 | vld1q_s16(in1 + 24) |
| 420 | }; |
| 421 | |
| 422 | const int16x8x4_t gy = |
| 423 | { |
| 424 | vld1q_s16(in2), |
| 425 | vld1q_s16(in2 + 8), |
| 426 | vld1q_s16(in2 + 16), |
| 427 | vld1q_s16(in2 + 24) |
| 428 | }; |
| 429 | |
| 430 | // Compute and store phase |
| 431 | vst1_u8(out2 + 0, phase_quantization_S16_S16(gx.val[0], gy.val[0])); |
| 432 | vst1_u8(out2 + 8, phase_quantization_S16_S16(gx.val[1], gy.val[1])); |
| 433 | vst1_u8(out2 + 16, phase_quantization_S16_S16(gx.val[2], gy.val[2])); |
| 434 | vst1_u8(out2 + 24, phase_quantization_S16_S16(gx.val[3], gy.val[3])); |
| 435 | |
| 436 | // Compute and store magnitude using L2 normalization |
| 437 | vst1q_u16(out1 + 0, mag_l2_S16_S16(gx.val[0], gy.val[0])); |
| 438 | vst1q_u16(out1 + 8, mag_l2_S16_S16(gx.val[1], gy.val[1])); |
| 439 | vst1q_u16(out1 + 16, mag_l2_S16_S16(gx.val[2], gy.val[2])); |
| 440 | vst1q_u16(out1 + 24, mag_l2_S16_S16(gx.val[3], gy.val[3])); |
| 441 | } |
| 442 | |
| 443 | /** Gradient function used when the gradient size = 7 and when the norm_type = L1-norm |
| 444 | * |
| 445 | * @param[in] in1_ptr Pointer to source image. Gx image. Data type supported S32 |
| 446 | * @param[in] in2_ptr Pointer to source image. Gy image. Data type supported S32 |
| 447 | * @param[out] out1_ptr Pointer to destination image. Magnitude. Data type supported U32 |
| 448 | * @param[out] out2_ptr Pointer to destination image. Quantized phase. Data type supported U8 |
| 449 | */ |
| 450 | void mag_phase_l1norm_S32_S32_U32_U8(const void *__restrict in1_ptr, const void *__restrict in2_ptr, void *__restrict out1_ptr, void *__restrict out2_ptr) |
| 451 | { |
| 452 | auto in1 = static_cast<const int32_t *__restrict>(in1_ptr); |
| 453 | auto in2 = static_cast<const int32_t *__restrict>(in2_ptr); |
| 454 | auto out1 = static_cast<uint32_t *__restrict>(out1_ptr); |
| 455 | auto out2 = static_cast<uint8_t *__restrict>(out2_ptr); |
| 456 | |
| 457 | // Process low and high part |
| 458 | for(size_t i = 0; i < 2; ++i, in1 += 16, in2 += 16, out1 += 16, out2 += 16) |
| 459 | { |
| 460 | const int32x4x2_t gx0 = |
| 461 | { |
| 462 | vld1q_s32(in1 + 0), |
| 463 | vld1q_s32(in1 + 4) |
| 464 | }; |
| 465 | |
| 466 | const int32x4x2_t gx1 = |
| 467 | { |
| 468 | vld1q_s32(in1 + 8), |
| 469 | vld1q_s32(in1 + 12) |
| 470 | }; |
| 471 | |
| 472 | const int32x4x2_t gy0 = |
| 473 | { |
| 474 | vld1q_s32(in2 + 0), |
| 475 | vld1q_s32(in2 + 4) |
| 476 | }; |
| 477 | |
| 478 | const int32x4x2_t gy1 = |
| 479 | { |
| 480 | vld1q_s32(in2 + 8), |
| 481 | vld1q_s32(in2 + 12) |
| 482 | }; |
| 483 | |
| 484 | // Compute and store phase |
| 485 | vst1_u8(out2 + 0, phase_quantization_S32_S32(gx0, gy0)); |
| 486 | vst1_u8(out2 + 8, phase_quantization_S32_S32(gx1, gy1)); |
| 487 | |
| 488 | // Compute magnitude using L1 normalization |
| 489 | const uint32x4x2_t mag0 = mag_l1_S32_S32(gx0, gy0); |
| 490 | const uint32x4x2_t mag1 = mag_l1_S32_S32(gx1, gy1); |
| 491 | |
| 492 | // Store magnitude |
| 493 | vst1q_u32(out1 + 0, mag0.val[0]); |
| 494 | vst1q_u32(out1 + 4, mag0.val[1]); |
| 495 | vst1q_u32(out1 + 8, mag1.val[0]); |
| 496 | vst1q_u32(out1 + 12, mag1.val[1]); |
| 497 | } |
| 498 | } |
| 499 | |
| 500 | /** Gradient function used when the gradient size = 7 and when the norm_type = L2-norm |
| 501 | * |
| 502 | * @param[in] in1_ptr Pointer to source image. Gx image. Data type supported S32 |
| 503 | * @param[in] in2_ptr Pointer to source image. Gy image. Data type supported S32 |
| 504 | * @param[out] out1_ptr Pointer to destination image. Magnitude. Data type supported U32 |
| 505 | * @param[out] out2_ptr Pointer to destination image. Quantized phase. Data type supported U8 |
| 506 | */ |
| 507 | void mag_phase_l2norm_S32_S32_U32_U8(const void *__restrict in1_ptr, const void *__restrict in2_ptr, void *__restrict out1_ptr, void *__restrict out2_ptr) |
| 508 | { |
| 509 | auto in1 = static_cast<const int32_t *__restrict>(in1_ptr); |
| 510 | auto in2 = static_cast<const int32_t *__restrict>(in2_ptr); |
| 511 | auto out1 = static_cast<uint32_t *__restrict>(out1_ptr); |
| 512 | auto out2 = static_cast<uint8_t *__restrict>(out2_ptr); |
| 513 | |
| 514 | // Process low and high part |
| 515 | for(size_t i = 0; i < 2; ++i, in1 += 16, in2 += 16, out1 += 16, out2 += 16) |
| 516 | { |
| 517 | const int32x4x2_t gx0 = |
| 518 | { |
| 519 | vld1q_s32(in1 + 0), |
| 520 | vld1q_s32(in1 + 4) |
| 521 | }; |
| 522 | |
| 523 | const int32x4x2_t gx1 = |
| 524 | { |
| 525 | vld1q_s32(in1 + 8), |
| 526 | vld1q_s32(in1 + 12) |
| 527 | }; |
| 528 | |
| 529 | const int32x4x2_t gy0 = |
| 530 | { |
| 531 | vld1q_s32(in2 + 0), |
| 532 | vld1q_s32(in2 + 4) |
| 533 | }; |
| 534 | |
| 535 | const int32x4x2_t gy1 = |
| 536 | { |
| 537 | vld1q_s32(in2 + 8), |
| 538 | vld1q_s32(in2 + 12) |
| 539 | }; |
| 540 | |
| 541 | // Compute and store phase |
| 542 | vst1_u8(out2 + 0, phase_quantization_S32_S32(gx0, gy0)); |
| 543 | vst1_u8(out2 + 8, phase_quantization_S32_S32(gx1, gy1)); |
| 544 | |
| 545 | // Compute magnitude using L2 normalization |
| 546 | const uint32x4x2_t mag0 = mag_l2_S32_S32(gx0, gy0); |
| 547 | const uint32x4x2_t mag1 = mag_l2_S32_S32(gx1, gy1); |
| 548 | |
| 549 | // Store magnitude |
| 550 | vst1q_u32(out1 + 0, mag0.val[0]); |
| 551 | vst1q_u32(out1 + 4, mag0.val[1]); |
| 552 | vst1q_u32(out1 + 8, mag1.val[0]); |
| 553 | vst1q_u32(out1 + 12, mag1.val[1]); |
| 554 | } |
| 555 | } |
| 556 | |
| 557 | inline uint16x4_t non_max_U32_helper(const uint32_t *in, const uint16x4_t pc, const uint32_t stride_mag, const int32_t lower_thr, const int32_t upper_thr) |
| 558 | { |
| 559 | // Phase for 4 pixel |
| 560 | const uint32x4_t pc32 = vmovl_u16(pc); |
| 561 | |
| 562 | // Get magnitude for 4 pixel |
| 563 | uint32x4_t mc = vld1q_u32(in); |
| 564 | |
| 565 | // Angle_quantized: 0 = 0°, 1 = 45°, 2 = 90°, 3 = 135° |
| 566 | // 0 degree |
| 567 | const uint32x4_t mk0_0 = vld1q_u32(in - 1); |
| 568 | const uint32x4_t mk0_1 = vld1q_u32(in + 1); |
| 569 | uint32x4_t mask0 = vceqq_u32(pc32, vdupq_n_u32(0)); |
Abe Mbise | 1b99338 | 2017-12-19 13:51:59 +0000 | [diff] [blame] | 570 | mask0 = vandq_u32(mask0, vcgtq_u32(mc, mk0_0)); |
| 571 | mask0 = vandq_u32(mask0, vcgtq_u32(mc, mk0_1)); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 572 | |
| 573 | // 45 degree |
| 574 | const uint32x4_t mk45_0 = vld1q_u32(in - stride_mag - 1); |
| 575 | const uint32x4_t mk45_1 = vld1q_u32(in + stride_mag + 1); |
| 576 | uint32x4_t mask1 = vceqq_u32(pc32, vdupq_n_u32(1)); |
Abe Mbise | 1b99338 | 2017-12-19 13:51:59 +0000 | [diff] [blame] | 577 | mask1 = vandq_u32(mask1, vcgtq_u32(mc, mk45_0)); |
| 578 | mask1 = vandq_u32(mask1, vcgtq_u32(mc, mk45_1)); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 579 | |
| 580 | // 90 degree |
| 581 | const uint32x4_t mk90_0 = vld1q_u32(in - stride_mag); |
| 582 | const uint32x4_t mk90_1 = vld1q_u32(in + stride_mag); |
| 583 | uint32x4_t mask2 = vceqq_u32(pc32, vdupq_n_u32(2)); |
Abe Mbise | 1b99338 | 2017-12-19 13:51:59 +0000 | [diff] [blame] | 584 | mask2 = vandq_u32(mask2, vcgtq_u32(mc, mk90_0)); |
| 585 | mask2 = vandq_u32(mask2, vcgtq_u32(mc, mk90_1)); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 586 | |
| 587 | // 135 degree |
| 588 | const uint32x4_t mk135_0 = vld1q_u32(in - stride_mag + 1); |
| 589 | const uint32x4_t mk135_1 = vld1q_u32(in + stride_mag - 1); |
| 590 | uint32x4_t mask3 = vceqq_u32(pc32, vdupq_n_u32(3)); |
Abe Mbise | 1b99338 | 2017-12-19 13:51:59 +0000 | [diff] [blame] | 591 | mask3 = vandq_u32(mask3, vcgtq_u32(mc, mk135_0)); |
| 592 | mask3 = vandq_u32(mask3, vcgtq_u32(mc, mk135_1)); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 593 | |
| 594 | // Merge masks |
| 595 | mask0 = vorrq_u32(mask0, mask1); |
| 596 | mask2 = vorrq_u32(mask2, mask3); |
| 597 | mask0 = vorrq_u32(mask0, mask2); |
| 598 | |
| 599 | mc = vbslq_u32(mask0, mc, vdupq_n_u32(0)); |
| 600 | |
| 601 | // mc > upper_thr |
| 602 | mask0 = vcgtq_u32(mc, vdupq_n_u32(upper_thr)); |
| 603 | |
| 604 | // mc <= lower_thr |
| 605 | mask1 = vcleq_u32(mc, vdupq_n_u32(lower_thr)); |
| 606 | |
| 607 | // mc <= upper_thr && mc > lower_thr |
| 608 | mask2 = vcleq_u32(mc, vdupq_n_u32(upper_thr)); |
| 609 | mask2 = vandq_u32(mask2, vcgtq_u32(mc, vdupq_n_u32(lower_thr))); |
| 610 | |
| 611 | mc = vbslq_u32(mask0, vdupq_n_u32(EDGE), mc); |
| 612 | mc = vbslq_u32(mask1, vdupq_n_u32(NO_EDGE), mc); |
| 613 | mc = vbslq_u32(mask2, vdupq_n_u32(MAYBE), mc); |
| 614 | |
| 615 | return vmovn_u32(mc); |
| 616 | } |
| 617 | |
| 618 | /** Computes edge tracing when is called by edge_trace_U8_U8 recursively |
| 619 | * |
| 620 | * @param[in] in Pointer to source image. Data type supported U8 |
| 621 | * @param[out] out Pointer to destination image. Data type supported U8 |
| 622 | * @param[in] in_stride Stride of the input image |
| 623 | * @param[in] out_stride Stride of the output image |
| 624 | */ |
| 625 | void edge_trace_recursive_U8_U8(uint8_t *__restrict in, uint8_t *__restrict out, const int32_t in_stride, const int32_t out_stride) |
| 626 | { |
| 627 | // Look for MAYBE pixels in 8 directions |
| 628 | *out = EDGE; |
| 629 | |
| 630 | // (-1, 0) |
| 631 | uint8_t pixel = *(in - 1); |
| 632 | |
| 633 | if(pixel == MAYBE) |
| 634 | { |
| 635 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 636 | *(in - 1) = EDGE; |
| 637 | |
| 638 | edge_trace_recursive_U8_U8(in - 1, out - 1, in_stride, out_stride); |
| 639 | } |
| 640 | |
| 641 | // (+1, 0) |
| 642 | pixel = *(in + 1); |
| 643 | |
| 644 | if(pixel == MAYBE) |
| 645 | { |
| 646 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 647 | *(in + 1) = EDGE; |
| 648 | |
| 649 | edge_trace_recursive_U8_U8(in + 1, out + 1, in_stride, out_stride); |
| 650 | } |
| 651 | |
| 652 | in -= in_stride; |
| 653 | out -= out_stride; |
| 654 | |
| 655 | // (-1, -1) |
| 656 | pixel = *(in - 1); |
| 657 | |
| 658 | if(pixel == MAYBE) |
| 659 | { |
| 660 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 661 | *(in - 1) = EDGE; |
| 662 | |
| 663 | edge_trace_recursive_U8_U8(in - 1, out - 1, in_stride, out_stride); |
| 664 | } |
| 665 | |
| 666 | // (0, -1) |
| 667 | pixel = *in; |
| 668 | |
| 669 | if(pixel == MAYBE) |
| 670 | { |
| 671 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 672 | *in = EDGE; |
| 673 | |
| 674 | edge_trace_recursive_U8_U8(in, out, in_stride, out_stride); |
| 675 | } |
| 676 | |
| 677 | // (+1, -1) |
| 678 | pixel = *(in + 1); |
| 679 | |
| 680 | if(pixel == MAYBE) |
| 681 | { |
| 682 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 683 | *(in + 1) = EDGE; |
| 684 | |
| 685 | edge_trace_recursive_U8_U8(in + 1, out + 1, in_stride, out_stride); |
| 686 | } |
| 687 | |
| 688 | in += in_stride * 2; |
| 689 | out += out_stride * 2; |
| 690 | |
| 691 | // (-1, +1) |
| 692 | pixel = *(in - 1); |
| 693 | |
| 694 | if(pixel == MAYBE) |
| 695 | { |
| 696 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 697 | *(in - 1) = EDGE; |
| 698 | |
| 699 | edge_trace_recursive_U8_U8(in - 1, out - 1, in_stride, out_stride); |
| 700 | } |
| 701 | |
| 702 | // (0, +1) |
| 703 | pixel = *in; |
| 704 | |
| 705 | if(pixel == MAYBE) |
| 706 | { |
| 707 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 708 | *in = EDGE; |
| 709 | |
| 710 | edge_trace_recursive_U8_U8(in, out, in_stride, out_stride); |
| 711 | } |
| 712 | |
| 713 | // (+1, +1) |
| 714 | pixel = *(in + 1); |
| 715 | |
| 716 | if(pixel == MAYBE) |
| 717 | { |
| 718 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 719 | *(in + 1) = EDGE; |
| 720 | |
| 721 | edge_trace_recursive_U8_U8(in + 1, out + 1, in_stride, out_stride); |
| 722 | } |
| 723 | } |
| 724 | } // namespace fp16 |
| 725 | |
| 726 | void NEGradientFP16Kernel::configure(const ITensor *gx, const ITensor *gy, ITensor *magnitude, ITensor *phase, int32_t norm_type) |
| 727 | { |
| 728 | ARM_COMPUTE_ERROR_ON_NULLPTR(gx, gy, magnitude, phase); |
| 729 | |
| 730 | set_shape_if_empty(*magnitude->info(), gx->info()->tensor_shape()); |
| 731 | set_shape_if_empty(*phase->info(), gx->info()->tensor_shape()); |
| 732 | |
| 733 | Format magnitude_format = gx->info()->data_type() == DataType::S16 ? Format::U16 : Format::U32; |
| 734 | set_format_if_unknown(*magnitude->info(), magnitude_format); |
| 735 | set_format_if_unknown(*phase->info(), Format::U8); |
| 736 | |
| 737 | ARM_COMPUTE_ERROR_ON_MISMATCHING_SHAPES(gx, gy, magnitude, phase); |
| 738 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(gx, 1, DataType::S16, DataType::S32); |
| 739 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(gy, 1, DataType::S16, DataType::S32); |
| 740 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(magnitude, 1, DataType::U16, DataType::U32); |
| 741 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(phase, 1, DataType::U8); |
| 742 | ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(gx, gy); |
| 743 | ARM_COMPUTE_ERROR_ON_MSG(element_size_from_data_type(gx->info()->data_type()) != element_size_from_data_type(magnitude->info()->data_type()), "Magnitude must have the same element size as Gx and Gy"); |
| 744 | |
| 745 | _gx = gx; |
| 746 | _gy = gy; |
| 747 | _magnitude = magnitude; |
| 748 | _phase = phase; |
| 749 | |
| 750 | if(_gx->info()->data_type() == DataType::S16) |
| 751 | { |
| 752 | if(norm_type == 1) |
| 753 | { |
| 754 | _func = &fp16::mag_phase_l1norm_S16_S16_U16_U8; |
| 755 | } |
| 756 | else |
| 757 | { |
| 758 | _func = &fp16::mag_phase_l2norm_S16_S16_U16_U8; |
| 759 | } |
| 760 | } |
| 761 | else |
| 762 | { |
| 763 | if(norm_type == 1) |
| 764 | { |
| 765 | _func = &fp16::mag_phase_l1norm_S32_S32_U32_U8; |
| 766 | } |
| 767 | else |
| 768 | { |
| 769 | _func = &fp16::mag_phase_l2norm_S32_S32_U32_U8; |
| 770 | } |
| 771 | } |
| 772 | |
| 773 | constexpr unsigned int num_elems_processed_per_iteration = 32; |
| 774 | |
| 775 | // Configure kernel window |
| 776 | Window win = calculate_max_window(*_gx->info(), Steps(num_elems_processed_per_iteration)); |
| 777 | |
| 778 | AccessWindowHorizontal gx_access(_gx->info(), 0, num_elems_processed_per_iteration); |
| 779 | AccessWindowHorizontal gy_access(_gy->info(), 0, num_elems_processed_per_iteration); |
| 780 | AccessWindowHorizontal mag_access(_magnitude->info(), 0, num_elems_processed_per_iteration); |
| 781 | AccessWindowHorizontal phase_access(_phase->info(), 0, num_elems_processed_per_iteration); |
| 782 | |
| 783 | update_window_and_padding(win, gx_access, gy_access, mag_access, phase_access); |
| 784 | |
| 785 | mag_access.set_valid_region(win, _gx->info()->valid_region()); |
| 786 | phase_access.set_valid_region(win, _gx->info()->valid_region()); |
| 787 | |
| 788 | INEKernel::configure(win); |
| 789 | } |
Ioan-Cristian Szabo | 5edbd1c | 2017-11-13 13:34:08 +0000 | [diff] [blame] | 790 | #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 791 | |
| 792 | namespace |
| 793 | { |
| 794 | inline uint8x8_t phase_quantization(const float32x4x2_t &gx, const float32x4x2_t &gy) |
| 795 | { |
| 796 | // Constant use for evaluating score1 and score3 |
| 797 | static const float32x4_t const45 = vdupq_n_f32(0.70710678118655f); |
| 798 | static const float32x4_t zero = vdupq_n_f32(0.0f); |
| 799 | static const float32x4_t one = vdupq_n_f32(1.0f); |
| 800 | static const float32x4_t two = vdupq_n_f32(2.0f); |
| 801 | static const float32x4_t three = vdupq_n_f32(3.0f); |
| 802 | |
| 803 | // Score0: (1, 0) |
| 804 | const float32x4x2_t score0 = |
| 805 | { |
| 806 | { |
| 807 | vabsq_f32(gx.val[0]), |
| 808 | vabsq_f32(gx.val[1]) |
| 809 | } |
| 810 | }; |
| 811 | |
| 812 | // Score2: ( 0, 1 ) |
| 813 | const float32x4x2_t score2 = |
| 814 | { |
| 815 | { |
| 816 | vabsq_f32(gy.val[0]), |
| 817 | vabsq_f32(gy.val[1]) |
| 818 | } |
| 819 | }; |
| 820 | |
| 821 | // Score1 and Score3: ( sqrt(2) / 2, sqrt(2) / 2 ) - ( -sqrt(2) / 2, sqrt(2) / 2 ) |
| 822 | float32x4x2_t score1 = |
| 823 | { |
| 824 | { |
| 825 | vmulq_f32(gy.val[0], const45), |
| 826 | vmulq_f32(gy.val[1], const45) |
| 827 | } |
| 828 | }; |
| 829 | |
| 830 | float32x4x2_t score3 = score1; |
| 831 | |
| 832 | score1.val[0] = vmlaq_f32(score1.val[0], gx.val[0], const45); |
| 833 | score1.val[1] = vmlaq_f32(score1.val[1], gx.val[1], const45); |
| 834 | score3.val[0] = vmlsq_f32(score3.val[0], gx.val[0], const45); |
| 835 | score3.val[1] = vmlsq_f32(score3.val[1], gx.val[1], const45); |
| 836 | |
| 837 | score1.val[0] = vabsq_f32(score1.val[0]); |
| 838 | score1.val[1] = vabsq_f32(score1.val[1]); |
| 839 | score3.val[0] = vabsq_f32(score3.val[0]); |
| 840 | score3.val[1] = vabsq_f32(score3.val[1]); |
| 841 | |
| 842 | float32x4x2_t phase = |
| 843 | { |
| 844 | { |
| 845 | zero, |
| 846 | zero |
| 847 | } |
| 848 | }; |
| 849 | |
| 850 | float32x4x2_t old_score = score0; |
| 851 | |
| 852 | // score1 > old_score? |
| 853 | uint32x4x2_t mask = |
| 854 | { |
| 855 | { |
| 856 | vcgtq_f32(score1.val[0], old_score.val[0]), |
| 857 | vcgtq_f32(score1.val[1], old_score.val[1]) |
| 858 | } |
| 859 | }; |
| 860 | |
| 861 | phase.val[0] = vbslq_f32(mask.val[0], one, phase.val[0]); |
| 862 | phase.val[1] = vbslq_f32(mask.val[1], one, phase.val[1]); |
| 863 | old_score.val[0] = vbslq_f32(mask.val[0], score1.val[0], old_score.val[0]); |
| 864 | old_score.val[1] = vbslq_f32(mask.val[1], score1.val[1], old_score.val[1]); |
| 865 | |
| 866 | // score2 > old_score? |
| 867 | mask.val[0] = vcgtq_f32(score2.val[0], old_score.val[0]); |
| 868 | mask.val[1] = vcgtq_f32(score2.val[1], old_score.val[1]); |
| 869 | |
| 870 | phase.val[0] = vbslq_f32(mask.val[0], two, phase.val[0]); |
| 871 | phase.val[1] = vbslq_f32(mask.val[1], two, phase.val[1]); |
| 872 | old_score.val[0] = vbslq_f32(mask.val[0], score2.val[0], old_score.val[0]); |
| 873 | old_score.val[1] = vbslq_f32(mask.val[1], score2.val[1], old_score.val[1]); |
| 874 | |
| 875 | // score3 > old_score? |
| 876 | mask.val[0] = vcgtq_f32(score3.val[0], old_score.val[0]); |
| 877 | mask.val[1] = vcgtq_f32(score3.val[1], old_score.val[1]); |
| 878 | |
| 879 | phase.val[0] = vbslq_f32(mask.val[0], three, phase.val[0]); |
| 880 | phase.val[1] = vbslq_f32(mask.val[1], three, phase.val[1]); |
| 881 | old_score.val[0] = vbslq_f32(mask.val[0], score3.val[0], old_score.val[0]); |
| 882 | old_score.val[1] = vbslq_f32(mask.val[1], score3.val[1], old_score.val[1]); |
| 883 | |
| 884 | // Convert from float32x4_t to uint8x8_t |
| 885 | return vmovn_u16(vcombine_u16(vmovn_u32(vcvtq_u32_f32(phase.val[0])), |
| 886 | vmovn_u32(vcvtq_u32_f32(phase.val[1])))); |
| 887 | } |
| 888 | |
| 889 | /* Computes the gradient phase if gradient_size = 3 or 5. The output is quantized. |
| 890 | * 0 = 0°, 1 = 45°, 2 = 90°, 3 = 135° |
| 891 | * |
| 892 | * @param[in] gx Gx component |
| 893 | * @param[in] gy Gy component |
| 894 | * |
| 895 | * @return quantized phase for 8 pixels |
| 896 | */ |
| 897 | inline uint8x8_t phase_quantization_S16_S16(int16x8_t gx, int16x8_t gy) |
| 898 | { |
| 899 | // Convert to float |
| 900 | const float32x4x2_t gx_f32 = |
| 901 | { |
| 902 | { |
| 903 | vcvtq_f32_s32(vmovl_s16(vget_low_s16(gx))), |
| 904 | vcvtq_f32_s32(vmovl_s16(vget_high_s16(gx))) |
| 905 | } |
| 906 | }; |
| 907 | |
| 908 | const float32x4x2_t gy_f32 = |
| 909 | { |
| 910 | { |
| 911 | vcvtq_f32_s32(vmovl_s16(vget_low_s16(gy))), |
| 912 | vcvtq_f32_s32(vmovl_s16(vget_high_s16(gy))) |
| 913 | } |
| 914 | }; |
| 915 | |
| 916 | return phase_quantization(gx_f32, gy_f32); |
| 917 | } |
| 918 | |
| 919 | /* Computes the gradient phase if gradient_size = 7. The output is quantized. |
| 920 | * 0 = 0°, 1 = 45°, 2 = 90°, 3 = 135° |
| 921 | * |
| 922 | * @param[in] gx Gx component |
| 923 | * @param[in] gy Gy component |
| 924 | * |
| 925 | * @return quantized phase for 8 pixels |
| 926 | */ |
| 927 | inline uint8x8_t phase_quantization_S32_S32(const int32x4x2_t &gx, const int32x4x2_t &gy) |
| 928 | { |
| 929 | // Convert to float |
| 930 | const float32x4x2_t gx_f32 = |
| 931 | { |
| 932 | { |
| 933 | vcvtq_f32_s32(gx.val[0]), |
| 934 | vcvtq_f32_s32(gx.val[1]) |
| 935 | } |
| 936 | }; |
| 937 | |
| 938 | const float32x4x2_t gy_f32 = |
| 939 | { |
| 940 | { |
| 941 | vcvtq_f32_s32(gy.val[0]), |
| 942 | vcvtq_f32_s32(gy.val[1]) |
| 943 | } |
| 944 | }; |
| 945 | |
| 946 | return phase_quantization(gx_f32, gy_f32); |
| 947 | } |
| 948 | |
| 949 | /* Computes the magnitude using the L1-norm type if gradient_size = 3 or 5 |
| 950 | * |
| 951 | * @param[in] gx Gx component |
| 952 | * @param[in] gy Gy component |
| 953 | * |
| 954 | * @return magnitude for 8 pixels |
| 955 | */ |
| 956 | inline uint16x8_t mag_l1_S16_S16(int16x8_t gx, int16x8_t gy) |
| 957 | { |
| 958 | return vaddq_u16(vreinterpretq_u16_s16(vabsq_s16(gx)), |
| 959 | vreinterpretq_u16_s16(vabsq_s16(gy))); |
| 960 | } |
| 961 | |
| 962 | /* Computes the magnitude using the L1-norm type if gradient_size = 7 |
| 963 | * |
| 964 | * @param[in] gx Gx component |
| 965 | * @param[in] gy Gy component |
| 966 | * |
| 967 | * @return magnitude for 8 pixels |
| 968 | */ |
| 969 | inline uint32x4x2_t mag_l1_S32_S32(const int32x4x2_t &gx, const int32x4x2_t &gy) |
| 970 | { |
| 971 | const uint32x4x2_t gx_abs = |
| 972 | { |
| 973 | { |
| 974 | vreinterpretq_u32_s32(vabsq_s32(gx.val[0])), |
| 975 | vreinterpretq_u32_s32(vabsq_s32(gx.val[1])) |
| 976 | } |
| 977 | }; |
| 978 | |
| 979 | const uint32x4x2_t gy_abs = |
| 980 | { |
| 981 | { |
| 982 | vreinterpretq_u32_s32(vabsq_s32(gy.val[0])), |
| 983 | vreinterpretq_u32_s32(vabsq_s32(gy.val[1])) |
| 984 | } |
| 985 | }; |
| 986 | |
| 987 | const uint32x4x2_t output = |
| 988 | { |
| 989 | { |
| 990 | vaddq_u32(gx_abs.val[0], gy_abs.val[0]), |
| 991 | vaddq_u32(gx_abs.val[1], gy_abs.val[1]) |
| 992 | } |
| 993 | }; |
| 994 | |
| 995 | return output; |
| 996 | } |
| 997 | |
| 998 | inline float32x4x2_t mag_l2(const float32x4x2_t &gx, const float32x4x2_t &gy) |
| 999 | { |
| 1000 | // x^2 ... |
| 1001 | float32x4x2_t magnitude = |
| 1002 | { |
| 1003 | { |
| 1004 | vmulq_f32(gx.val[0], gx.val[0]), |
| 1005 | vmulq_f32(gx.val[1], gx.val[1]) |
| 1006 | } |
| 1007 | }; |
| 1008 | |
| 1009 | // ... + y^2 |
| 1010 | magnitude.val[0] = vmlaq_f32(magnitude.val[0], gy.val[0], gy.val[0]); |
| 1011 | magnitude.val[1] = vmlaq_f32(magnitude.val[1], gy.val[1], gy.val[1]); |
| 1012 | |
| 1013 | // sqrt(...) |
| 1014 | magnitude.val[0] = vmulq_f32(vrsqrteq_f32(magnitude.val[0]), magnitude.val[0]); |
| 1015 | magnitude.val[1] = vmulq_f32(vrsqrteq_f32(magnitude.val[1]), magnitude.val[1]); |
| 1016 | |
| 1017 | return magnitude; |
| 1018 | } |
| 1019 | |
| 1020 | /* Computes the magnitude using L2-norm if gradient_size = 3 or 5 |
| 1021 | * |
| 1022 | * @param[in] gx Gx component |
| 1023 | * @param[in] gy Gy component |
| 1024 | * |
| 1025 | * @return magnitude for 8 pixels |
| 1026 | */ |
| 1027 | inline uint16x8_t mag_l2_S16_S16(int16x8_t gx, int16x8_t gy) |
| 1028 | { |
| 1029 | // Compute magnitude using L2 normalization |
| 1030 | const float32x4x2_t gx2 = |
| 1031 | { |
| 1032 | { |
| 1033 | vcvtq_f32_s32(vmovl_s16(vget_low_s16(gx))), |
| 1034 | vcvtq_f32_s32(vmovl_s16(vget_high_s16(gx))) |
| 1035 | } |
| 1036 | }; |
| 1037 | |
| 1038 | const float32x4x2_t gy2 = |
| 1039 | { |
| 1040 | { |
| 1041 | vcvtq_f32_s32(vmovl_s16(vget_low_s16(gy))), |
| 1042 | vcvtq_f32_s32(vmovl_s16(vget_high_s16(gy))) |
| 1043 | } |
| 1044 | }; |
| 1045 | |
| 1046 | const float32x4x2_t magnitude = mag_l2(gx2, gy2); |
| 1047 | |
| 1048 | // Store magnitude - Convert to uint16x8 |
| 1049 | return vcombine_u16(vmovn_u32(vcvtq_u32_f32(magnitude.val[0])), |
| 1050 | vmovn_u32(vcvtq_u32_f32(magnitude.val[1]))); |
| 1051 | } |
| 1052 | |
| 1053 | /* Computes the magnitude using L2-norm if gradient_size = 7 |
| 1054 | * |
| 1055 | * @param[in] gx Gx component |
| 1056 | * @param[in] gy Gy component |
| 1057 | * |
| 1058 | * @return magnitude for 8 pixels |
| 1059 | */ |
| 1060 | inline uint32x4x2_t mag_l2_S32_S32(const int32x4x2_t &gx, const int32x4x2_t &gy) |
| 1061 | { |
| 1062 | // Compute magnitude using L2 normalization |
| 1063 | float32x4x2_t gx2 = |
| 1064 | { |
| 1065 | { |
| 1066 | vcvtq_f32_s32(gx.val[0]), |
| 1067 | vcvtq_f32_s32(gx.val[1]) |
| 1068 | } |
| 1069 | }; |
| 1070 | |
| 1071 | float32x4x2_t gy2 = |
| 1072 | { |
| 1073 | { |
| 1074 | vcvtq_f32_s32(gy.val[0]), |
| 1075 | vcvtq_f32_s32(gy.val[1]) |
| 1076 | } |
| 1077 | }; |
| 1078 | |
| 1079 | const float32x4x2_t magnitude = mag_l2(gx2, gy2); |
| 1080 | const uint32x4x2_t mag32 = |
| 1081 | { |
| 1082 | { |
| 1083 | vcvtq_u32_f32(magnitude.val[0]), |
| 1084 | vcvtq_u32_f32(magnitude.val[1]) |
| 1085 | } |
| 1086 | }; |
| 1087 | |
| 1088 | return mag32; |
| 1089 | } |
| 1090 | |
| 1091 | /* Gradient function used when the gradient size = 3 or 5 and when the norm_type = L1-norm |
| 1092 | * |
| 1093 | * @param[in] gx_ptr Pointer to source image. Gx image. Data type supported S16 |
| 1094 | * @param[in] gy_ptr Pointer to source image. Gy image. Data type supported S16 |
| 1095 | * @param[out] magnitude_ptr Pointer to destination image. Magnitude. Data type supported U16 |
| 1096 | * @param[out] phase_ptr Pointer to destination image. Quantized phase. Data type supported U8 |
| 1097 | */ |
| 1098 | void mag_phase_l1norm_S16_S16_U16_U8(const void *__restrict gx_ptr, const void *__restrict gy_ptr, void *__restrict magnitude_ptr, void *__restrict phase_ptr) |
| 1099 | { |
| 1100 | const auto gx = static_cast<const int16_t *__restrict>(gx_ptr); |
| 1101 | const auto gy = static_cast<const int16_t *__restrict>(gy_ptr); |
| 1102 | const auto magnitude = static_cast<uint16_t *__restrict>(magnitude_ptr); |
| 1103 | const auto phase = static_cast<uint8_t *__restrict>(phase_ptr); |
| 1104 | |
| 1105 | const int16x8x4_t gx_val = |
| 1106 | { |
| 1107 | { |
| 1108 | vld1q_s16(gx), |
| 1109 | vld1q_s16(gx + 8), |
| 1110 | vld1q_s16(gx + 16), |
| 1111 | vld1q_s16(gx + 24) |
| 1112 | } |
| 1113 | }; |
| 1114 | |
| 1115 | const int16x8x4_t gy_val = |
| 1116 | { |
| 1117 | { |
| 1118 | vld1q_s16(gy), |
| 1119 | vld1q_s16(gy + 8), |
| 1120 | vld1q_s16(gy + 16), |
| 1121 | vld1q_s16(gy + 24) |
| 1122 | } |
| 1123 | }; |
| 1124 | |
| 1125 | // Compute and store phase |
| 1126 | vst1_u8(phase + 0, phase_quantization_S16_S16(gx_val.val[0], gy_val.val[0])); |
| 1127 | vst1_u8(phase + 8, phase_quantization_S16_S16(gx_val.val[1], gy_val.val[1])); |
| 1128 | vst1_u8(phase + 16, phase_quantization_S16_S16(gx_val.val[2], gy_val.val[2])); |
| 1129 | vst1_u8(phase + 24, phase_quantization_S16_S16(gx_val.val[3], gy_val.val[3])); |
| 1130 | |
| 1131 | // Compute ans store magnitude using L1 normalization |
| 1132 | vst1q_u16(magnitude + 0, mag_l1_S16_S16(gx_val.val[0], gy_val.val[0])); |
| 1133 | vst1q_u16(magnitude + 8, mag_l1_S16_S16(gx_val.val[1], gy_val.val[1])); |
| 1134 | vst1q_u16(magnitude + 16, mag_l1_S16_S16(gx_val.val[2], gy_val.val[2])); |
| 1135 | vst1q_u16(magnitude + 24, mag_l1_S16_S16(gx_val.val[3], gy_val.val[3])); |
| 1136 | } |
| 1137 | |
| 1138 | /* Gradient function used when the gradient size = 3 or 5 and when the norm_type = L2-norm |
| 1139 | * |
| 1140 | * @param[in] gx_ptr Pointer to source image. Gx image. Data type supported S16 |
| 1141 | * @param[in] gy_ptr Pointer to source image. Gy image. Data type supported S16 |
| 1142 | * @param[out] magnitude_ptr Pointer to destination image. Magnitude. Data type supported U16 |
| 1143 | * @param[out] phase_ptr Pointer to destination image. Quantized phase. Data type supported U8 |
| 1144 | */ |
| 1145 | void mag_phase_l2norm_S16_S16_U16_U8(const void *__restrict gx_ptr, const void *__restrict gy_ptr, void *__restrict magnitude_ptr, void *__restrict phase_ptr) |
| 1146 | { |
| 1147 | const auto gx = static_cast<const int16_t *__restrict>(gx_ptr); |
| 1148 | const auto gy = static_cast<const int16_t *__restrict>(gy_ptr); |
| 1149 | const auto magnitude = static_cast<uint16_t *__restrict>(magnitude_ptr); |
| 1150 | const auto phase = static_cast<uint8_t *__restrict>(phase_ptr); |
| 1151 | |
| 1152 | const int16x8x4_t gx_val = |
| 1153 | { |
| 1154 | { |
| 1155 | vld1q_s16(gx), |
| 1156 | vld1q_s16(gx + 8), |
| 1157 | vld1q_s16(gx + 16), |
| 1158 | vld1q_s16(gx + 24) |
| 1159 | } |
| 1160 | }; |
| 1161 | |
| 1162 | const int16x8x4_t gy_val = |
| 1163 | { |
| 1164 | { |
| 1165 | vld1q_s16(gy), |
| 1166 | vld1q_s16(gy + 8), |
| 1167 | vld1q_s16(gy + 16), |
| 1168 | vld1q_s16(gy + 24) |
| 1169 | } |
| 1170 | }; |
| 1171 | |
| 1172 | // Compute and store phase |
| 1173 | vst1_u8(phase + 0, phase_quantization_S16_S16(gx_val.val[0], gy_val.val[0])); |
| 1174 | vst1_u8(phase + 8, phase_quantization_S16_S16(gx_val.val[1], gy_val.val[1])); |
| 1175 | vst1_u8(phase + 16, phase_quantization_S16_S16(gx_val.val[2], gy_val.val[2])); |
| 1176 | vst1_u8(phase + 24, phase_quantization_S16_S16(gx_val.val[3], gy_val.val[3])); |
| 1177 | |
| 1178 | // Compute and store magnitude using L2 normalization |
| 1179 | vst1q_u16(magnitude + 0, mag_l2_S16_S16(gx_val.val[0], gy_val.val[0])); |
| 1180 | vst1q_u16(magnitude + 8, mag_l2_S16_S16(gx_val.val[1], gy_val.val[1])); |
| 1181 | vst1q_u16(magnitude + 16, mag_l2_S16_S16(gx_val.val[2], gy_val.val[2])); |
| 1182 | vst1q_u16(magnitude + 24, mag_l2_S16_S16(gx_val.val[3], gy_val.val[3])); |
| 1183 | } |
| 1184 | |
| 1185 | /* Gradient function used when the gradient size = 7 and when the norm_type = L1-norm |
| 1186 | * |
| 1187 | * @param[in] gx_ptr Pointer to source image. Gx image. Data type supported S32 |
| 1188 | * @param[in] gy_ptr Pointer to source image. Gy image. Data type supported S32 |
| 1189 | * @param[out] magnitude_ptr Pointer to destination image. Magnitude. Data type supported U32 |
| 1190 | * @param[out] phase_ptr Pointer to destination image. Quantized phase. Data type support U8 |
| 1191 | */ |
| 1192 | void mag_phase_l1norm_S32_S32_U32_U8(const void *__restrict gx_ptr, const void *__restrict gy_ptr, void *__restrict magnitude_ptr, void *__restrict phase_ptr) |
| 1193 | { |
| 1194 | auto gx = static_cast<const int32_t *__restrict>(gx_ptr); |
| 1195 | auto gy = static_cast<const int32_t *__restrict>(gy_ptr); |
| 1196 | auto magnitude = static_cast<uint32_t *__restrict>(magnitude_ptr); |
| 1197 | auto phase = static_cast<uint8_t *__restrict>(phase_ptr); |
| 1198 | |
| 1199 | // Process low and high part |
| 1200 | for(size_t i = 0; i < 2; ++i, gx += 16, gy += 16, magnitude += 16, phase += 16) |
| 1201 | { |
| 1202 | const int32x4x2_t gx0 = |
| 1203 | { |
| 1204 | { |
| 1205 | vld1q_s32(gx + 0), |
| 1206 | vld1q_s32(gx + 4) |
| 1207 | } |
| 1208 | }; |
| 1209 | |
| 1210 | const int32x4x2_t gx1 = |
| 1211 | { |
| 1212 | { |
| 1213 | vld1q_s32(gx + 8), |
| 1214 | vld1q_s32(gx + 12) |
| 1215 | } |
| 1216 | }; |
| 1217 | |
| 1218 | const int32x4x2_t gy0 = |
| 1219 | { |
| 1220 | { |
| 1221 | vld1q_s32(gy + 0), |
| 1222 | vld1q_s32(gy + 4) |
| 1223 | } |
| 1224 | }; |
| 1225 | |
| 1226 | const int32x4x2_t gy1 = |
| 1227 | { |
| 1228 | { |
| 1229 | vld1q_s32(gy + 8), |
| 1230 | vld1q_s32(gy + 12) |
| 1231 | } |
| 1232 | }; |
| 1233 | |
| 1234 | // Compute and store phase |
| 1235 | vst1_u8(phase + 0, phase_quantization_S32_S32(gx0, gy0)); |
| 1236 | vst1_u8(phase + 8, phase_quantization_S32_S32(gx1, gy1)); |
| 1237 | |
| 1238 | // Compute magnitude using L1 normalization |
| 1239 | const uint32x4x2_t mag0 = mag_l1_S32_S32(gx0, gy0); |
| 1240 | const uint32x4x2_t mag1 = mag_l1_S32_S32(gx1, gy1); |
| 1241 | |
| 1242 | // Store magnitude |
| 1243 | vst1q_u32(magnitude + 0, mag0.val[0]); |
| 1244 | vst1q_u32(magnitude + 4, mag0.val[1]); |
| 1245 | vst1q_u32(magnitude + 8, mag1.val[0]); |
| 1246 | vst1q_u32(magnitude + 12, mag1.val[1]); |
| 1247 | } |
| 1248 | } |
| 1249 | |
| 1250 | /* Gradient function used when the gradient size = 7 and when the norm_type = L2-norm |
| 1251 | * |
| 1252 | * @param[in] gx_ptr Pointer to source image. Gx image. Data type supported S32 |
| 1253 | * @param[in] gy_ptr Pointer to source image. Gy image. Data type supported S32 |
| 1254 | * @param[out] magnitude_ptr Pointer to destination image. Magnitude. Data type supported U32 |
| 1255 | * @param[out] phase_ptr Pointer to destination image. Quantized phase. Data type supported U8 |
| 1256 | */ |
| 1257 | void mag_phase_l2norm_S32_S32_U32_U8(const void *__restrict gx_ptr, const void *__restrict gy_ptr, void *__restrict magnitude_ptr, void *__restrict phase_ptr) |
| 1258 | { |
| 1259 | auto gx = static_cast<const int32_t *__restrict>(gx_ptr); |
| 1260 | auto gy = static_cast<const int32_t *__restrict>(gy_ptr); |
| 1261 | auto magnitude = static_cast<uint32_t *__restrict>(magnitude_ptr); |
| 1262 | auto phase = static_cast<uint8_t *__restrict>(phase_ptr); |
| 1263 | |
| 1264 | // Process low and high part |
| 1265 | for(size_t i = 0; i < 2; ++i, gx += 16, gy += 16, magnitude += 16, phase += 16) |
| 1266 | { |
| 1267 | const int32x4x2_t gx0 = |
| 1268 | { |
| 1269 | { |
| 1270 | vld1q_s32(gx + 0), |
| 1271 | vld1q_s32(gx + 4) |
| 1272 | } |
| 1273 | }; |
| 1274 | |
| 1275 | const int32x4x2_t gx1 = |
| 1276 | { |
| 1277 | { |
| 1278 | vld1q_s32(gx + 8), |
| 1279 | vld1q_s32(gx + 12) |
| 1280 | } |
| 1281 | }; |
| 1282 | |
| 1283 | const int32x4x2_t gy0 = |
| 1284 | { |
| 1285 | { |
| 1286 | vld1q_s32(gy + 0), |
| 1287 | vld1q_s32(gy + 4) |
| 1288 | } |
| 1289 | }; |
| 1290 | |
| 1291 | const int32x4x2_t gy1 = |
| 1292 | { |
| 1293 | { |
| 1294 | vld1q_s32(gy + 8), |
| 1295 | vld1q_s32(gy + 12) |
| 1296 | } |
| 1297 | }; |
| 1298 | |
| 1299 | // Compute and store phase |
| 1300 | vst1_u8(phase + 0, phase_quantization_S32_S32(gx0, gy0)); |
| 1301 | vst1_u8(phase + 8, phase_quantization_S32_S32(gx1, gy1)); |
| 1302 | |
| 1303 | // Compute magnitude using L2 normalization |
| 1304 | const uint32x4x2_t mag0 = mag_l2_S32_S32(gx0, gy0); |
| 1305 | const uint32x4x2_t mag1 = mag_l2_S32_S32(gx1, gy1); |
| 1306 | |
| 1307 | // Store magnitude |
| 1308 | vst1q_u32(magnitude + 0, mag0.val[0]); |
| 1309 | vst1q_u32(magnitude + 4, mag0.val[1]); |
| 1310 | vst1q_u32(magnitude + 8, mag1.val[0]); |
| 1311 | vst1q_u32(magnitude + 12, mag1.val[1]); |
| 1312 | } |
| 1313 | } |
| 1314 | |
| 1315 | /* Computes non-maxima suppression and hysteresis when the gradient size = 3 or 5 |
| 1316 | * |
| 1317 | * @param[in] magnitude_ptr Pointer to source image. Magnitude. Data type supported U16 |
| 1318 | * @param[in] phase_ptr Pointer to source image. Quantized phase. Data type supported U8 |
| 1319 | * @param[out] output_ptr Pointer to output image. Data type supported U8 |
| 1320 | * @param[in] stride_mag Stride of magnitude image |
| 1321 | * @param[in] lower_thr Lower threshold used for the hysteresis |
| 1322 | * @param[in] upper_thr Upper threshold used for the hysteresis |
| 1323 | */ |
| 1324 | void non_max_suppression_U16_U8_U8(const void *__restrict magnitude_ptr, const void *__restrict phase_ptr, void *__restrict output_ptr, const uint32_t stride_mag, const int32_t lower_thr, |
| 1325 | const int32_t upper_thr) |
| 1326 | { |
| 1327 | const auto magnitude = static_cast<const uint16_t *__restrict>(magnitude_ptr); |
| 1328 | const auto phase = static_cast<const uint8_t *__restrict>(phase_ptr); |
| 1329 | const auto output = static_cast<uint8_t *__restrict>(output_ptr); |
| 1330 | |
| 1331 | // Get magnitude and phase of the centre pixels |
| 1332 | uint16x8_t mc = vld1q_u16(magnitude); |
| 1333 | |
| 1334 | // Angle_quantized: 0 = 0°, 1 = 45°, 2 = 90°, 3 = 135° |
| 1335 | const uint16x8_t pc16 = vmovl_u8(vld1_u8(phase)); |
| 1336 | |
| 1337 | // 0 degree |
| 1338 | const uint16x8_t mk0_0 = vld1q_u16(magnitude - 1); |
| 1339 | const uint16x8_t mk0_1 = vld1q_u16(magnitude + 1); |
| 1340 | uint16x8_t mask0 = vceqq_u16(pc16, vdupq_n_u16(0)); |
Abe Mbise | 1b99338 | 2017-12-19 13:51:59 +0000 | [diff] [blame] | 1341 | mask0 = vandq_u16(mask0, vcgtq_u16(mc, mk0_0)); |
| 1342 | mask0 = vandq_u16(mask0, vcgtq_u16(mc, mk0_1)); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1343 | |
| 1344 | // 45 degree |
| 1345 | const uint16x8_t mk45_0 = vld1q_u16(magnitude - stride_mag - 1); |
| 1346 | const uint16x8_t mk45_1 = vld1q_u16(magnitude + stride_mag + 1); |
| 1347 | uint16x8_t mask1 = vceqq_u16(pc16, vdupq_n_u16(1)); |
Abe Mbise | 1b99338 | 2017-12-19 13:51:59 +0000 | [diff] [blame] | 1348 | mask1 = vandq_u16(mask1, vcgtq_u16(mc, mk45_0)); |
| 1349 | mask1 = vandq_u16(mask1, vcgtq_u16(mc, mk45_1)); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1350 | |
| 1351 | // 90 degree |
| 1352 | const uint16x8_t mk90_0 = vld1q_u16(magnitude - stride_mag); |
| 1353 | const uint16x8_t mk90_1 = vld1q_u16(magnitude + stride_mag); |
| 1354 | uint16x8_t mask2 = vceqq_u16(pc16, vdupq_n_u16(2)); |
Abe Mbise | 1b99338 | 2017-12-19 13:51:59 +0000 | [diff] [blame] | 1355 | mask2 = vandq_u16(mask2, vcgtq_u16(mc, mk90_0)); |
| 1356 | mask2 = vandq_u16(mask2, vcgtq_u16(mc, mk90_1)); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1357 | |
| 1358 | // 135 degree |
| 1359 | const uint16x8_t mk135_0 = vld1q_u16(magnitude - stride_mag + 1); |
| 1360 | const uint16x8_t mk135_1 = vld1q_u16(magnitude + stride_mag - 1); |
| 1361 | uint16x8_t mask3 = vceqq_u16(pc16, vdupq_n_u16(3)); |
Abe Mbise | 1b99338 | 2017-12-19 13:51:59 +0000 | [diff] [blame] | 1362 | mask3 = vandq_u16(mask3, vcgtq_u16(mc, mk135_0)); |
| 1363 | mask3 = vandq_u16(mask3, vcgtq_u16(mc, mk135_1)); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1364 | |
| 1365 | // Merge masks |
| 1366 | mask0 = vorrq_u16(mask0, mask1); |
| 1367 | mask2 = vorrq_u16(mask2, mask3); |
| 1368 | mask0 = vorrq_u16(mask0, mask2); |
| 1369 | |
| 1370 | mc = vbslq_u16(mask0, mc, vdupq_n_u16(0)); |
| 1371 | |
| 1372 | // mc > upper_thr |
| 1373 | mask0 = vcgtq_u16(mc, vdupq_n_u16(upper_thr)); |
| 1374 | |
| 1375 | // mc <= lower_thr |
| 1376 | mask1 = vcleq_u16(mc, vdupq_n_u16(lower_thr)); |
| 1377 | |
| 1378 | // mc <= upper_thr && mc > lower_thr |
| 1379 | mask2 = vcleq_u16(mc, vdupq_n_u16(upper_thr)); |
| 1380 | mask2 = vandq_u16(mask2, vcgtq_u16(mc, vdupq_n_u16(lower_thr))); |
| 1381 | |
| 1382 | mc = vbslq_u16(mask0, vdupq_n_u16(EDGE), mc); |
| 1383 | mc = vbslq_u16(mask1, vdupq_n_u16(NO_EDGE), mc); |
| 1384 | mc = vbslq_u16(mask2, vdupq_n_u16(MAYBE), mc); |
| 1385 | |
| 1386 | vst1_u8(output, vmovn_u16(mc)); |
| 1387 | } |
| 1388 | |
| 1389 | inline uint16x4_t non_max_U32_helper(const uint32_t *input, const uint16x4_t pc, const uint32_t stride_mag, const int32_t lower_thr, const int32_t upper_thr) |
| 1390 | { |
| 1391 | // Phase for 4 pixel |
| 1392 | const uint32x4_t pc32 = vmovl_u16(pc); |
| 1393 | |
| 1394 | // Get magnitude for 4 pixel |
| 1395 | uint32x4_t mc = vld1q_u32(input); |
| 1396 | |
| 1397 | // Angle_quantized: 0 = 0°, 1 = 45°, 2 = 90°, 3 = 135° |
| 1398 | // 0 degree |
| 1399 | const uint32x4_t mk0_0 = vld1q_u32(input - 1); |
| 1400 | const uint32x4_t mk0_1 = vld1q_u32(input + 1); |
| 1401 | uint32x4_t mask0 = vceqq_u32(pc32, vdupq_n_u32(0)); |
Abe Mbise | 1b99338 | 2017-12-19 13:51:59 +0000 | [diff] [blame] | 1402 | mask0 = vandq_u32(mask0, vcgtq_u32(mc, mk0_0)); |
| 1403 | mask0 = vandq_u32(mask0, vcgtq_u32(mc, mk0_1)); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1404 | |
| 1405 | // 45 degree |
| 1406 | const uint32x4_t mk45_0 = vld1q_u32(input - stride_mag - 1); |
| 1407 | const uint32x4_t mk45_1 = vld1q_u32(input + stride_mag + 1); |
| 1408 | uint32x4_t mask1 = vceqq_u32(pc32, vdupq_n_u32(1)); |
Abe Mbise | 1b99338 | 2017-12-19 13:51:59 +0000 | [diff] [blame] | 1409 | mask1 = vandq_u32(mask1, vcgtq_u32(mc, mk45_0)); |
| 1410 | mask1 = vandq_u32(mask1, vcgtq_u32(mc, mk45_1)); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1411 | |
| 1412 | // 90 degree |
| 1413 | const uint32x4_t mk90_0 = vld1q_u32(input - stride_mag); |
| 1414 | const uint32x4_t mk90_1 = vld1q_u32(input + stride_mag); |
| 1415 | uint32x4_t mask2 = vceqq_u32(pc32, vdupq_n_u32(2)); |
Abe Mbise | 1b99338 | 2017-12-19 13:51:59 +0000 | [diff] [blame] | 1416 | mask2 = vandq_u32(mask2, vcgtq_u32(mc, mk90_0)); |
| 1417 | mask2 = vandq_u32(mask2, vcgtq_u32(mc, mk90_1)); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1418 | |
| 1419 | // 135 degree |
| 1420 | const uint32x4_t mk135_0 = vld1q_u32(input - stride_mag + 1); |
| 1421 | const uint32x4_t mk135_1 = vld1q_u32(input + stride_mag - 1); |
| 1422 | uint32x4_t mask3 = vceqq_u32(pc32, vdupq_n_u32(3)); |
Abe Mbise | 1b99338 | 2017-12-19 13:51:59 +0000 | [diff] [blame] | 1423 | mask3 = vandq_u32(mask3, vcgtq_u32(mc, mk135_0)); |
| 1424 | mask3 = vandq_u32(mask3, vcgtq_u32(mc, mk135_1)); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1425 | |
| 1426 | // Merge masks |
| 1427 | mask0 = vorrq_u32(mask0, mask1); |
| 1428 | mask2 = vorrq_u32(mask2, mask3); |
| 1429 | mask0 = vorrq_u32(mask0, mask2); |
| 1430 | |
| 1431 | mc = vbslq_u32(mask0, mc, vdupq_n_u32(0)); |
| 1432 | |
| 1433 | // mc > upper_thr |
| 1434 | mask0 = vcgtq_u32(mc, vdupq_n_u32(upper_thr)); |
| 1435 | |
| 1436 | // mc <= lower_thr |
| 1437 | mask1 = vcleq_u32(mc, vdupq_n_u32(lower_thr)); |
| 1438 | |
| 1439 | // mc <= upper_thr && mc > lower_thr |
| 1440 | mask2 = vcleq_u32(mc, vdupq_n_u32(upper_thr)); |
| 1441 | mask2 = vandq_u32(mask2, vcgtq_u32(mc, vdupq_n_u32(lower_thr))); |
| 1442 | |
| 1443 | mc = vbslq_u32(mask0, vdupq_n_u32(EDGE), mc); |
| 1444 | mc = vbslq_u32(mask1, vdupq_n_u32(NO_EDGE), mc); |
| 1445 | mc = vbslq_u32(mask2, vdupq_n_u32(MAYBE), mc); |
| 1446 | |
| 1447 | return vmovn_u32(mc); |
| 1448 | } |
| 1449 | |
| 1450 | /* Computes non-maxima suppression and hysteresis when the gradient_size = 7 |
| 1451 | * |
| 1452 | * @param[in] magnitude_ptr Pointer to source image. Magnitude. Data type supported U32 |
| 1453 | * @param[in] phase_ptr Pointer to source image. Quantized phase. Data type supported U8 |
| 1454 | * @param[out] output_ptr Pointer to destination image. Data type supported U8 |
| 1455 | * @param[in] stride_mag Stride of magnitude image |
| 1456 | * @param[in] lower_thr Lower threshold used for the hysteresis |
| 1457 | * @param[in] upper_thr Upper threshold used for the hysteresis |
| 1458 | */ |
| 1459 | void non_max_suppression_U32_U8_U8(const void *__restrict magnitude_ptr, const void *__restrict phase_ptr, void *__restrict output_ptr, const uint32_t stride_mag, const int32_t lower_thr, |
| 1460 | const int32_t upper_thr) |
| 1461 | { |
| 1462 | const auto magnitude = static_cast<const uint32_t *__restrict>(magnitude_ptr); |
| 1463 | const auto phase = static_cast<const uint8_t *__restrict>(phase_ptr); |
| 1464 | const auto output = static_cast<uint8_t *__restrict>(output_ptr); |
| 1465 | |
| 1466 | // Get phase for 8 pixel |
| 1467 | const uint16x8_t pc16 = vmovl_u8(vld1_u8(phase)); |
| 1468 | |
| 1469 | // Compute non maxima suppression |
| 1470 | const uint16x4x2_t res = |
| 1471 | { |
| 1472 | { |
| 1473 | non_max_U32_helper(magnitude, vget_low_u16(pc16), stride_mag, lower_thr, upper_thr), |
| 1474 | non_max_U32_helper(magnitude + 4, vget_high_u16(pc16), stride_mag, lower_thr, upper_thr) |
| 1475 | } |
| 1476 | }; |
| 1477 | |
| 1478 | // Store result |
| 1479 | vst1_u8(output, vmovn_u16(vcombine_u16(res.val[0], res.val[1]))); |
| 1480 | } |
| 1481 | |
| 1482 | /* Computes edge tracing when is called by edge_trace_U8_U8 recursively |
| 1483 | * |
| 1484 | * @param[in] input Pointer to source image. Data type supported U8 |
| 1485 | * @param[out] output Pointer to destination image. Data type supported U8 |
| 1486 | * @param[in] input_stride Stride of the input image |
| 1487 | * @param[in] output_stride Stride of the output image |
| 1488 | */ |
| 1489 | void edge_trace_recursive_U8_U8(uint8_t *__restrict input, uint8_t *__restrict output, const int32_t input_stride, const int32_t output_stride) |
| 1490 | { |
| 1491 | // Look for MAYBE pixels in 8 directions |
| 1492 | *output = EDGE; |
| 1493 | |
| 1494 | // (-1, 0) |
| 1495 | uint8_t pixel = *(input - 1); |
| 1496 | |
| 1497 | if(pixel == MAYBE) |
| 1498 | { |
| 1499 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 1500 | *(input - 1) = EDGE; |
| 1501 | |
| 1502 | edge_trace_recursive_U8_U8(input - 1, output - 1, input_stride, output_stride); |
| 1503 | } |
| 1504 | |
| 1505 | // (+1, 0) |
| 1506 | pixel = *(input + 1); |
| 1507 | |
| 1508 | if(pixel == MAYBE) |
| 1509 | { |
| 1510 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 1511 | *(input + 1) = EDGE; |
| 1512 | |
| 1513 | edge_trace_recursive_U8_U8(input + 1, output + 1, input_stride, output_stride); |
| 1514 | } |
| 1515 | |
| 1516 | input -= input_stride; |
| 1517 | output -= output_stride; |
| 1518 | |
| 1519 | // (-1, -1) |
| 1520 | pixel = *(input - 1); |
| 1521 | |
| 1522 | if(pixel == MAYBE) |
| 1523 | { |
| 1524 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 1525 | *(input - 1) = EDGE; |
| 1526 | |
| 1527 | edge_trace_recursive_U8_U8(input - 1, output - 1, input_stride, output_stride); |
| 1528 | } |
| 1529 | |
| 1530 | // (0, -1) |
| 1531 | pixel = *input; |
| 1532 | |
| 1533 | if(pixel == MAYBE) |
| 1534 | { |
| 1535 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 1536 | *input = EDGE; |
| 1537 | |
| 1538 | edge_trace_recursive_U8_U8(input, output, input_stride, output_stride); |
| 1539 | } |
| 1540 | |
| 1541 | // (+1, -1) |
| 1542 | pixel = *(input + 1); |
| 1543 | |
| 1544 | if(pixel == MAYBE) |
| 1545 | { |
| 1546 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 1547 | *(input + 1) = EDGE; |
| 1548 | |
| 1549 | edge_trace_recursive_U8_U8(input + 1, output + 1, input_stride, output_stride); |
| 1550 | } |
| 1551 | |
| 1552 | input += input_stride * 2; |
| 1553 | output += output_stride * 2; |
| 1554 | |
| 1555 | // (-1, +1) |
| 1556 | pixel = *(input - 1); |
| 1557 | |
| 1558 | if(pixel == MAYBE) |
| 1559 | { |
| 1560 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 1561 | *(input - 1) = EDGE; |
| 1562 | |
| 1563 | edge_trace_recursive_U8_U8(input - 1, output - 1, input_stride, output_stride); |
| 1564 | } |
| 1565 | |
| 1566 | // (0, +1) |
| 1567 | pixel = *input; |
| 1568 | |
| 1569 | if(pixel == MAYBE) |
| 1570 | { |
| 1571 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 1572 | *input = EDGE; |
| 1573 | |
| 1574 | edge_trace_recursive_U8_U8(input, output, input_stride, output_stride); |
| 1575 | } |
| 1576 | |
| 1577 | // (+1, +1) |
| 1578 | pixel = *(input + 1); |
| 1579 | |
| 1580 | if(pixel == MAYBE) |
| 1581 | { |
| 1582 | // Touched a MAYBE point. MAYBE becomes EDGE |
| 1583 | *(input + 1) = EDGE; |
| 1584 | |
| 1585 | edge_trace_recursive_U8_U8(input + 1, output + 1, input_stride, output_stride); |
| 1586 | } |
| 1587 | } |
| 1588 | |
| 1589 | /* Computes edge tracing |
| 1590 | * |
| 1591 | * @param[in] input Pointer to source image. Data type supported U8 |
| 1592 | * @param[out] output Pointer to destination image. Data type supported U8 |
| 1593 | * @param[in] input_stride Stride of the input image |
| 1594 | * @param[in] output_stride Stride of the output image |
| 1595 | */ |
| 1596 | void edge_trace_U8_U8(uint8_t *__restrict input, uint8_t *__restrict output, const int32_t input_stride, const int32_t output_stride) |
| 1597 | { |
| 1598 | if(*input == NO_EDGE) |
| 1599 | { |
| 1600 | *output = NO_EDGE; |
| 1601 | } |
| 1602 | // Check if EDGE and not yet touched |
| 1603 | else if((*input == EDGE) && (*output == NO_EDGE)) |
| 1604 | { |
| 1605 | edge_trace_recursive_U8_U8(input, output, input_stride, output_stride); |
| 1606 | } |
| 1607 | } |
| 1608 | } // namespace |
| 1609 | |
| 1610 | NEGradientKernel::NEGradientKernel() |
| 1611 | : _func(nullptr), _gx(nullptr), _gy(nullptr), _magnitude(nullptr), _phase(nullptr) |
| 1612 | { |
| 1613 | } |
| 1614 | |
| 1615 | void NEGradientKernel::configure(const ITensor *gx, const ITensor *gy, ITensor *magnitude, ITensor *phase, int32_t norm_type) |
| 1616 | { |
| 1617 | ARM_COMPUTE_ERROR_ON_NULLPTR(gx, gy, magnitude, phase); |
| 1618 | |
| 1619 | set_shape_if_empty(*magnitude->info(), gx->info()->tensor_shape()); |
| 1620 | set_shape_if_empty(*phase->info(), gx->info()->tensor_shape()); |
| 1621 | |
| 1622 | Format magnitude_format = gx->info()->data_type() == DataType::S16 ? Format::U16 : Format::U32; |
| 1623 | set_format_if_unknown(*magnitude->info(), magnitude_format); |
| 1624 | set_format_if_unknown(*phase->info(), Format::U8); |
| 1625 | |
| 1626 | ARM_COMPUTE_ERROR_ON_MISMATCHING_SHAPES(gx, gy, magnitude, phase); |
| 1627 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(gx, 1, DataType::S16, DataType::S32); |
| 1628 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(gy, 1, DataType::S16, DataType::S32); |
| 1629 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(magnitude, 1, DataType::U16, DataType::U32); |
| 1630 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(phase, 1, DataType::U8); |
| 1631 | ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(gx, gy); |
| 1632 | ARM_COMPUTE_ERROR_ON_MSG(element_size_from_data_type(gx->info()->data_type()) != element_size_from_data_type(magnitude->info()->data_type()), "Magnitude must have the same element size as Gx and Gy"); |
| 1633 | |
| 1634 | _gx = gx; |
| 1635 | _gy = gy; |
| 1636 | _magnitude = magnitude; |
| 1637 | _phase = phase; |
| 1638 | |
| 1639 | if(_gx->info()->data_type() == DataType::S16) |
| 1640 | { |
| 1641 | if(norm_type == 1) |
| 1642 | { |
| 1643 | _func = &mag_phase_l1norm_S16_S16_U16_U8; |
| 1644 | } |
| 1645 | else |
| 1646 | { |
| 1647 | _func = &mag_phase_l2norm_S16_S16_U16_U8; |
| 1648 | } |
| 1649 | } |
| 1650 | else |
| 1651 | { |
| 1652 | if(norm_type == 1) |
| 1653 | { |
| 1654 | _func = &mag_phase_l1norm_S32_S32_U32_U8; |
| 1655 | } |
| 1656 | else |
| 1657 | { |
| 1658 | _func = &mag_phase_l2norm_S32_S32_U32_U8; |
| 1659 | } |
| 1660 | } |
| 1661 | |
| 1662 | constexpr unsigned int num_elems_processed_per_iteration = 32; |
| 1663 | |
| 1664 | // Configure kernel window |
| 1665 | Window win = calculate_max_window(*_gx->info(), Steps(num_elems_processed_per_iteration)); |
| 1666 | |
| 1667 | AccessWindowHorizontal gx_access(_gx->info(), 0, num_elems_processed_per_iteration); |
| 1668 | AccessWindowHorizontal gy_access(_gy->info(), 0, num_elems_processed_per_iteration); |
| 1669 | AccessWindowHorizontal mag_access(_magnitude->info(), 0, num_elems_processed_per_iteration); |
| 1670 | AccessWindowHorizontal phase_access(_phase->info(), 0, num_elems_processed_per_iteration); |
| 1671 | |
| 1672 | update_window_and_padding(win, gx_access, gy_access, mag_access, phase_access); |
| 1673 | |
| 1674 | mag_access.set_valid_region(win, _gx->info()->valid_region()); |
| 1675 | phase_access.set_valid_region(win, _gx->info()->valid_region()); |
| 1676 | |
| 1677 | INEKernel::configure(win); |
| 1678 | } |
| 1679 | |
Moritz Pflanzer | c186b57 | 2017-09-07 09:48:04 +0100 | [diff] [blame] | 1680 | void NEGradientKernel::run(const Window &window, const ThreadInfo &info) |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1681 | { |
Moritz Pflanzer | c186b57 | 2017-09-07 09:48:04 +0100 | [diff] [blame] | 1682 | ARM_COMPUTE_UNUSED(info); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1683 | ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); |
| 1684 | ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window); |
| 1685 | ARM_COMPUTE_ERROR_ON(_func == nullptr); |
| 1686 | Iterator gx(_gx, window); |
| 1687 | Iterator gy(_gy, window); |
| 1688 | Iterator magnitude(_magnitude, window); |
| 1689 | Iterator phase(_phase, window); |
| 1690 | |
| 1691 | execute_window_loop(window, [&](const Coordinates & id) |
| 1692 | { |
| 1693 | (*_func)(gx.ptr(), gy.ptr(), magnitude.ptr(), phase.ptr()); |
| 1694 | }, |
| 1695 | gx, gy, magnitude, phase); |
| 1696 | } |
| 1697 | |
| 1698 | NEEdgeNonMaxSuppressionKernel::NEEdgeNonMaxSuppressionKernel() |
| 1699 | : _func(nullptr), _magnitude(nullptr), _phase(nullptr), _output(nullptr), _lower_thr(0), _upper_thr(0) |
| 1700 | { |
| 1701 | } |
| 1702 | |
| 1703 | BorderSize NEEdgeNonMaxSuppressionKernel::border_size() const |
| 1704 | { |
| 1705 | return BorderSize(1); |
| 1706 | } |
| 1707 | |
| 1708 | void NEEdgeNonMaxSuppressionKernel::configure(const ITensor *magnitude, const ITensor *phase, ITensor *output, |
| 1709 | int32_t upper_thr, int32_t lower_thr, bool border_undefined) |
| 1710 | { |
| 1711 | ARM_COMPUTE_ERROR_ON_NULLPTR(magnitude, phase, output); |
| 1712 | |
| 1713 | set_shape_if_empty(*output->info(), magnitude->info()->tensor_shape()); |
| 1714 | |
| 1715 | set_format_if_unknown(*phase->info(), Format::U8); |
| 1716 | set_format_if_unknown(*output->info(), Format::U8); |
| 1717 | |
| 1718 | ARM_COMPUTE_ERROR_ON_MISMATCHING_SHAPES(magnitude, phase, output); |
| 1719 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(magnitude, 1, DataType::U16, DataType::U32); |
| 1720 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(phase, 1, DataType::U8); |
| 1721 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8); |
| 1722 | ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(phase, output); |
| 1723 | |
| 1724 | _magnitude = magnitude; |
| 1725 | _phase = phase; |
| 1726 | _output = output; |
| 1727 | |
| 1728 | switch(_magnitude->info()->data_type()) |
| 1729 | { |
| 1730 | case DataType::U16: |
| 1731 | _func = &non_max_suppression_U16_U8_U8; |
| 1732 | break; |
| 1733 | case DataType::U32: |
| 1734 | _func = &non_max_suppression_U32_U8_U8; |
| 1735 | break; |
| 1736 | default: |
| 1737 | ARM_COMPUTE_ERROR("Unsupported data type!"); |
| 1738 | } |
| 1739 | |
| 1740 | // Set thresholds |
| 1741 | _lower_thr = lower_thr; |
| 1742 | _upper_thr = upper_thr; |
| 1743 | |
| 1744 | constexpr unsigned int num_elems_processed_per_iteration = 8; |
| 1745 | constexpr unsigned int num_elems_read_per_iteration = 10; |
| 1746 | constexpr unsigned int num_rows_read_per_iteration = 3; |
| 1747 | |
| 1748 | // Configure kernel window |
| 1749 | Window win = calculate_max_window(*_magnitude->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size()); |
| 1750 | |
| 1751 | AccessWindowRectangle mag_access(_magnitude->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration); |
| 1752 | AccessWindowHorizontal phase_access(_phase->info(), 0, num_elems_processed_per_iteration); |
| 1753 | AccessWindowHorizontal output_access(_output->info(), 0, num_elems_processed_per_iteration); |
| 1754 | |
| 1755 | update_window_and_padding(win, mag_access, phase_access, output_access); |
| 1756 | |
| 1757 | output_access.set_valid_region(win, _magnitude->info()->valid_region(), border_undefined, border_size()); |
| 1758 | |
| 1759 | INEKernel::configure(win); |
| 1760 | } |
| 1761 | |
Moritz Pflanzer | c186b57 | 2017-09-07 09:48:04 +0100 | [diff] [blame] | 1762 | void NEEdgeNonMaxSuppressionKernel::run(const Window &window, const ThreadInfo &info) |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1763 | { |
Moritz Pflanzer | c186b57 | 2017-09-07 09:48:04 +0100 | [diff] [blame] | 1764 | ARM_COMPUTE_UNUSED(info); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1765 | ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); |
| 1766 | ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window); |
| 1767 | ARM_COMPUTE_ERROR_ON(_func == nullptr); |
| 1768 | Iterator magnitude(_magnitude, window); |
| 1769 | Iterator phase(_phase, window); |
| 1770 | Iterator output(_output, window); |
| 1771 | |
| 1772 | const size_t input1_stride = _magnitude->info()->strides_in_bytes()[1]; |
| 1773 | const size_t input1_stride_ushort = input1_stride / data_size_from_type(_magnitude->info()->data_type()); |
| 1774 | |
| 1775 | execute_window_loop(window, [&](const Coordinates & id) |
| 1776 | { |
| 1777 | (*_func)(magnitude.ptr(), phase.ptr(), output.ptr(), input1_stride_ushort, _lower_thr, _upper_thr); |
| 1778 | }, |
| 1779 | magnitude, phase, output); |
| 1780 | } |
| 1781 | |
| 1782 | NEEdgeTraceKernel::NEEdgeTraceKernel() |
| 1783 | : _input(nullptr), _output(nullptr) |
| 1784 | { |
| 1785 | } |
| 1786 | |
| 1787 | BorderSize NEEdgeTraceKernel::border_size() const |
| 1788 | { |
| 1789 | return BorderSize(1); |
| 1790 | } |
| 1791 | |
| 1792 | bool NEEdgeTraceKernel::is_parallelisable() const |
| 1793 | { |
| 1794 | return false; |
| 1795 | } |
| 1796 | |
| 1797 | void NEEdgeTraceKernel::configure(ITensor *input, ITensor *output) |
| 1798 | { |
| 1799 | ARM_COMPUTE_ERROR_ON_NULLPTR(input, output); |
| 1800 | |
| 1801 | set_shape_if_empty(*output->info(), input->info()->tensor_shape()); |
| 1802 | |
| 1803 | set_format_if_unknown(*input->info(), Format::U8); |
| 1804 | set_format_if_unknown(*output->info(), Format::U8); |
| 1805 | |
| 1806 | ARM_COMPUTE_ERROR_ON_MISMATCHING_SHAPES(input, output); |
| 1807 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8); |
| 1808 | ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8); |
| 1809 | ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output); |
| 1810 | |
| 1811 | _input = input; |
| 1812 | _output = output; |
| 1813 | |
| 1814 | constexpr unsigned int num_elems_processed_per_iteration = 1; |
| 1815 | |
| 1816 | // Configure kernel window |
| 1817 | Window win = calculate_max_window(*_input->info(), Steps(num_elems_processed_per_iteration)); |
| 1818 | |
| 1819 | const ValidRegion &input_valid_region = input->info()->valid_region(); |
| 1820 | const ValidRegion &output_valid_region = output->info()->valid_region(); |
| 1821 | |
| 1822 | // Reads can occur within the valid region of the input + border |
| 1823 | AccessWindowStatic input_access(input->info(), |
| 1824 | input_valid_region.anchor[0] - border_size().left, |
| 1825 | input_valid_region.anchor[1] - border_size().top, |
| 1826 | input_valid_region.anchor[0] + input_valid_region.shape[0] + border_size().right, |
| 1827 | input_valid_region.anchor[1] + input_valid_region.shape[1] + border_size().bottom); |
| 1828 | |
| 1829 | // Writes can occur within the valid region of the output + border |
| 1830 | AccessWindowStatic output_access(output->info(), |
| 1831 | output_valid_region.anchor[0] - border_size().left, |
| 1832 | output_valid_region.anchor[1] - border_size().top, |
| 1833 | output_valid_region.anchor[0] + output_valid_region.shape[0] + border_size().right, |
| 1834 | output_valid_region.anchor[1] + output_valid_region.shape[1] + border_size().bottom); |
| 1835 | |
| 1836 | update_window_and_padding(win, input_access, output_access); |
| 1837 | |
| 1838 | output_access.set_valid_region(win, _input->info()->valid_region()); |
| 1839 | |
| 1840 | INEKernel::configure(win); |
| 1841 | } |
| 1842 | |
Moritz Pflanzer | c186b57 | 2017-09-07 09:48:04 +0100 | [diff] [blame] | 1843 | void NEEdgeTraceKernel::run(const Window &window, const ThreadInfo &info) |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1844 | { |
Moritz Pflanzer | c186b57 | 2017-09-07 09:48:04 +0100 | [diff] [blame] | 1845 | ARM_COMPUTE_UNUSED(info); |
Anthony Barbier | 6ff3b19 | 2017-09-04 18:44:23 +0100 | [diff] [blame] | 1846 | ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); |
| 1847 | ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window); |
| 1848 | Iterator input(_input, window); |
| 1849 | Iterator output(_output, window); |
| 1850 | |
| 1851 | const size_t input_stride = _input->info()->strides_in_bytes()[1]; |
| 1852 | const size_t output_stride = _output->info()->strides_in_bytes()[1]; |
| 1853 | |
| 1854 | execute_window_loop(window, [&](const Coordinates & id) |
| 1855 | { |
| 1856 | edge_trace_U8_U8(input.ptr(), output.ptr(), input_stride, output_stride); |
| 1857 | }, |
| 1858 | input, output); |
| 1859 | } |