1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
/*
* By downloading, copying, installing or using the software you agree to this license.
* If you do not agree to this license, do not download, install,
* copy or use the software.
*
*
* License Agreement
* For Open Source Computer Vision Library
* (3-clause BSD License)
*
* Copyright (C) 2015, NVIDIA Corporation, all rights reserved.
* Third party copyrights are property of their respective owners.
*
* Redistribution and use in source and binary forms, with or without modification,
* are permitted provided that the following conditions are met:
*
* * Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
*
* * Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
*
* * Neither the names of the copyright holders nor the names of the contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* This software is provided by the copyright holders and contributors "as is" and
* any express or implied warranties, including, but not limited to, the implied
* warranties of merchantability and fitness for a particular purpose are disclaimed.
* In no event shall copyright holders or contributors be liable for any direct,
* indirect, incidental, special, exemplary, or consequential damages
* (including, but not limited to, procurement of substitute goods or services;
* loss of use, data, or profits; or business interruption) however caused
* and on any theory of liability, whether in contract, strict liability,
* or tort (including negligence or otherwise) arising in any way out of
* the use of this software, even if advised of the possibility of such damage.
*/
#include "remap.hpp"
namespace CAROTENE_NS {
bool isWarpAffineNearestNeighborSupported(const Size2D &ssize)
{
#if SIZE_MAX > UINT32_MAX
return !(ssize.width > 0xffffFFFF || ssize.height > 0xffffFFFF) && // Restrict image size since internal index evaluation
// is performed with u32
isSupportedConfiguration();
#else
(void)ssize;
return isSupportedConfiguration();
#endif
}
bool isWarpAffineLinearSupported(const Size2D &ssize)
{
#if SIZE_MAX > UINT32_MAX
return !(ssize.width > 0xffffFFFF || ssize.height > 0xffffFFFF) && // Restrict image size since internal index evaluation
// is performed with u32
isSupportedConfiguration();
#else
(void)ssize;
return isSupportedConfiguration();
#endif
}
void warpAffineNearestNeighbor(const Size2D &ssize, const Size2D &dsize,
const u8 * srcBase, ptrdiff_t srcStride,
const f32 * m,
u8 * dstBase, ptrdiff_t dstStride,
BORDER_MODE borderMode, u8 borderValue)
{
internal::assertSupportedConfiguration(isWarpAffineNearestNeighborSupported(ssize));
#ifdef CAROTENE_NEON
using namespace internal;
s32 _map[BLOCK_SIZE * BLOCK_SIZE + 16];
s32 * map = alignPtr(_map, 16);
int32x4_t v_width4 = vdupq_n_s32(ssize.width - 1), v_height4 = vdupq_n_s32(ssize.height - 1);
int32x4_t v_step4 = vdupq_n_s32(srcStride);
float32x4_t v_4 = vdupq_n_f32(4.0f);
float32x4_t v_m0 = vdupq_n_f32(m[0]);
float32x4_t v_m1 = vdupq_n_f32(m[1]);
float32x4_t v_m2 = vdupq_n_f32(m[2]);
float32x4_t v_m3 = vdupq_n_f32(m[3]);
float32x4_t v_m4 = vdupq_n_f32(m[4]);
float32x4_t v_m5 = vdupq_n_f32(m[5]);
if (borderMode == BORDER_MODE_REPLICATE)
{
int32x4_t v_zero4 = vdupq_n_s32(0);
for (size_t i = 0; i < dsize.height; i += BLOCK_SIZE)
{
size_t blockHeight = std::min<size_t>(BLOCK_SIZE, dsize.height - i);
for (size_t j = 0; j < dsize.width; j += BLOCK_SIZE)
{
size_t blockWidth = std::min<size_t>(BLOCK_SIZE, dsize.width - j);
// compute table
for (size_t y = 0; y < blockHeight; ++y)
{
s32 * map_row = getRowPtr(&map[0], blockWidth * sizeof(s32), y);
size_t x = 0, y_ = y + i;
f32 indeces[4] = { j + 0.0f, j + 1.0f, j + 2.0f, j + 3.0f };
float32x4_t v_x = vld1q_f32(indeces), v_y = vdupq_n_f32(y_);
float32x4_t v_yx = vmlaq_f32(v_m4, v_m2, v_y), v_yy = vmlaq_f32(v_m5, v_m3, v_y);
for ( ; x + 4 <= blockWidth; x += 4)
{
float32x4_t v_src_xf = vmlaq_f32(v_yx, v_m0, v_x);
float32x4_t v_src_yf = vmlaq_f32(v_yy, v_m1, v_x);
int32x4_t v_src_x = vmaxq_s32(v_zero4, vminq_s32(v_width4, vcvtq_s32_f32(v_src_xf)));
int32x4_t v_src_y = vmaxq_s32(v_zero4, vminq_s32(v_height4, vcvtq_s32_f32(v_src_yf)));
int32x4_t v_src_index = vmlaq_s32(v_src_x, v_src_y, v_step4);
vst1q_s32(map_row + x, v_src_index);
v_x = vaddq_f32(v_x, v_4);
}
f32 yx = m[2] * y_ + m[4], yy = m[3] * y_ + m[5];
for (ptrdiff_t x_ = x + j; x < blockWidth; ++x, ++x_)
{
f32 src_x_f = m[0] * x_ + yx;
f32 src_y_f = m[1] * x_ + yy;
s32 src_x = floorf(src_x_f), src_y = floorf(src_y_f);
src_x = std::max(0, std::min<s32>(ssize.width - 1, src_x));
src_y = std::max(0, std::min<s32>(ssize.height - 1, src_y));
map_row[x] = src_y * srcStride + src_x;
}
}
// make remap
remapNearestNeighborReplicate(Size2D(blockWidth, blockHeight), srcBase, &map[0],
getRowPtr(dstBase, dstStride, i) + j, dstStride);
}
}
}
else if (borderMode == BORDER_MODE_CONSTANT)
{
int32x4_t v_m1_4 = vdupq_n_s32(-1);
float32x4_t v_zero4 = vdupq_n_f32(0.0f);
for (size_t i = 0; i < dsize.height; i += BLOCK_SIZE)
{
size_t blockHeight = std::min<size_t>(BLOCK_SIZE, dsize.height - i);
for (size_t j = 0; j < dsize.width; j += BLOCK_SIZE)
{
size_t blockWidth = std::min<size_t>(BLOCK_SIZE, dsize.width - j);
// compute table
for (size_t y = 0; y < blockHeight; ++y)
{
s32 * map_row = getRowPtr(&map[0], blockWidth * sizeof(s32), y);
size_t x = 0, y_ = y + i;
f32 indeces[4] = { j + 0.0f, j + 1.0f, j + 2.0f, j + 3.0f };
float32x4_t v_x = vld1q_f32(indeces), v_y = vdupq_n_f32(y_);
float32x4_t v_yx = vmlaq_f32(v_m4, v_m2, v_y), v_yy = vmlaq_f32(v_m5, v_m3, v_y);
for ( ; x + 4 <= blockWidth; x += 4)
{
float32x4_t v_src_xf = vmlaq_f32(v_yx, v_m0, v_x);
float32x4_t v_src_yf = vmlaq_f32(v_yy, v_m1, v_x);
int32x4_t v_src_x = vcvtq_s32_f32(v_src_xf);
int32x4_t v_src_y = vcvtq_s32_f32(v_src_yf);
uint32x4_t v_mask = vandq_u32(vandq_u32(vcgeq_f32(v_src_xf, v_zero4), vcleq_s32(v_src_x, v_width4)),
vandq_u32(vcgeq_f32(v_src_yf, v_zero4), vcleq_s32(v_src_y, v_height4)));
int32x4_t v_src_index = vbslq_s32(v_mask, vmlaq_s32(v_src_x, v_src_y, v_step4), v_m1_4);
vst1q_s32(map_row + x, v_src_index);
v_x = vaddq_f32(v_x, v_4);
}
f32 yx = m[2] * y_ + m[4], yy = m[3] * y_ + m[5];
for (ptrdiff_t x_ = x + j; x < blockWidth; ++x, ++x_)
{
f32 src_x_f = m[0] * x_ + yx;
f32 src_y_f = m[1] * x_ + yy;
s32 src_x = floorf(src_x_f), src_y = floorf(src_y_f);
map_row[x] = (src_x >= 0) && (src_x < (s32)ssize.width) &&
(src_y >= 0) && (src_y < (s32)ssize.height) ? src_y * srcStride + src_x : -1;
}
}
// make remap
remapNearestNeighborConst(Size2D(blockWidth, blockHeight), srcBase, &map[0],
getRowPtr(dstBase, dstStride, i) + j, dstStride, borderValue);
}
}
}
#else
(void)ssize;
(void)dsize;
(void)srcBase;
(void)srcStride;
(void)m;
(void)dstBase;
(void)dstStride;
(void)borderMode;
(void)borderValue;
#endif
}
void warpAffineLinear(const Size2D &ssize, const Size2D &dsize,
const u8 * srcBase, ptrdiff_t srcStride,
const f32 * m,
u8 * dstBase, ptrdiff_t dstStride,
BORDER_MODE borderMode, u8 borderValue)
{
internal::assertSupportedConfiguration(isWarpAffineLinearSupported(ssize));
#ifdef CAROTENE_NEON
using namespace internal;
s32 _map[((BLOCK_SIZE * BLOCK_SIZE) << 2) + 16];
f32 _coeffs[((BLOCK_SIZE * BLOCK_SIZE) << 1) + 16];
s32 * map = alignPtr(_map, 16);
f32 * coeffs = alignPtr(_coeffs, 16);
int32x4_t v_width4 = vdupq_n_s32(ssize.width - 1), v_height4 = vdupq_n_s32(ssize.height - 1);
int32x4_t v_step4 = vdupq_n_s32(srcStride), v_1 = vdupq_n_s32(1);
float32x4_t v_zero4f = vdupq_n_f32(0.0f), v_one4f = vdupq_n_f32(1.0f);
float32x4_t v_m0 = vdupq_n_f32(m[0]);
float32x4_t v_m1 = vdupq_n_f32(m[1]);
float32x4_t v_m2 = vdupq_n_f32(m[2]);
float32x4_t v_m3 = vdupq_n_f32(m[3]);
float32x4_t v_m4 = vdupq_n_f32(m[4]);
float32x4_t v_m5 = vdupq_n_f32(m[5]);
if (borderMode == BORDER_MODE_REPLICATE)
{
int32x4_t v_zero4 = vdupq_n_s32(0);
for (size_t i = 0; i < dsize.height; i += BLOCK_SIZE)
{
size_t blockHeight = std::min<size_t>(BLOCK_SIZE, dsize.height - i);
for (size_t j = 0; j < dsize.width; j += BLOCK_SIZE)
{
size_t blockWidth = std::min<size_t>(BLOCK_SIZE, dsize.width - j);
// compute table
for (size_t y = 0; y < blockHeight; ++y)
{
s32 * map_row = getRowPtr(map, blockWidth * sizeof(s32) * 4, y);
f32 * coeff_row = getRowPtr(coeffs, blockWidth * sizeof(f32) * 2, y);
size_t x = 0, y_ = y + i;
f32 indeces[4] = { j + 0.0f, j + 1.0f, j + 2.0f, j + 3.0f };
float32x4_t v_x = vld1q_f32(indeces), v_y = vdupq_n_f32(y_), v_4 = vdupq_n_f32(4.0f);
float32x4_t v_yx = vmlaq_f32(v_m4, v_m2, v_y), v_yy = vmlaq_f32(v_m5, v_m3, v_y);
for ( ; x + 4 <= blockWidth; x += 4)
{
float32x4_t v_src_xf = vmlaq_f32(v_yx, v_m0, v_x);
float32x4_t v_src_yf = vmlaq_f32(v_yy, v_m1, v_x);
int32x4_t v_src_x = vcvtq_s32_f32(v_src_xf);
int32x4_t v_src_y = vcvtq_s32_f32(v_src_yf);
float32x4x2_t v_coeff;
v_coeff.val[0] = vsubq_f32(v_src_xf, vcvtq_f32_s32(v_src_x));
v_coeff.val[1] = vsubq_f32(v_src_yf, vcvtq_f32_s32(v_src_y));
uint32x4_t v_maskx = vcltq_f32(v_coeff.val[0], v_zero4f);
uint32x4_t v_masky = vcltq_f32(v_coeff.val[1], v_zero4f);
v_coeff.val[0] = vbslq_f32(v_maskx, vaddq_f32(v_one4f, v_coeff.val[0]), v_coeff.val[0]);
v_coeff.val[1] = vbslq_f32(v_masky, vaddq_f32(v_one4f, v_coeff.val[1]), v_coeff.val[1]);
v_src_x = vbslq_s32(v_maskx, vsubq_s32(v_src_x, v_1), v_src_x);
v_src_y = vbslq_s32(v_masky, vsubq_s32(v_src_y, v_1), v_src_y);
int32x4_t v_dst0_x = vmaxq_s32(v_zero4, vminq_s32(v_width4, v_src_x));
int32x4_t v_dst0_y = vmaxq_s32(v_zero4, vminq_s32(v_height4, v_src_y));
int32x4_t v_dst1_x = vmaxq_s32(v_zero4, vminq_s32(v_width4, vaddq_s32(v_1, v_src_x)));
int32x4_t v_dst1_y = vmaxq_s32(v_zero4, vminq_s32(v_height4, vaddq_s32(v_1, v_src_y)));
int32x4x4_t v_dst_index;
v_dst_index.val[0] = vmlaq_s32(v_dst0_x, v_dst0_y, v_step4);
v_dst_index.val[1] = vmlaq_s32(v_dst1_x, v_dst0_y, v_step4);
v_dst_index.val[2] = vmlaq_s32(v_dst0_x, v_dst1_y, v_step4);
v_dst_index.val[3] = vmlaq_s32(v_dst1_x, v_dst1_y, v_step4);
vst2q_f32(coeff_row + (x << 1), v_coeff);
vst4q_s32(map_row + (x << 2), v_dst_index);
v_x = vaddq_f32(v_x, v_4);
}
f32 yx = m[2] * y_ + m[4], yy = m[3] * y_ + m[5];
for (ptrdiff_t x_ = x + j; x < blockWidth; ++x, ++x_)
{
f32 src_x_f = m[0] * x_ + yx;
f32 src_y_f = m[1] * x_ + yy;
s32 src0_x = (s32)floorf(src_x_f);
s32 src0_y = (s32)floorf(src_y_f);
coeff_row[(x << 1) + 0] = src_x_f - src0_x;
coeff_row[(x << 1) + 1] = src_y_f - src0_y;
s32 src1_y = std::max(0, std::min<s32>(ssize.height - 1, src0_y + 1));
src0_y = std::max(0, std::min<s32>(ssize.height - 1, src0_y));
s32 src1_x = std::max(0, std::min<s32>(ssize.width - 1, src0_x + 1));
src0_x = std::max(0, std::min<s32>(ssize.width - 1, src0_x));
map_row[(x << 2) + 0] = src0_y * srcStride + src0_x;
map_row[(x << 2) + 1] = src0_y * srcStride + src1_x;
map_row[(x << 2) + 2] = src1_y * srcStride + src0_x;
map_row[(x << 2) + 3] = src1_y * srcStride + src1_x;
}
}
remapLinearReplicate(Size2D(blockWidth, blockHeight),
srcBase, &map[0], &coeffs[0],
getRowPtr(dstBase, dstStride, i) + j, dstStride);
}
}
}
else if (borderMode == BORDER_MODE_CONSTANT)
{
float32x4_t v_zero4 = vdupq_n_f32(0.0f);
int32x4_t v_m1_4 = vdupq_n_s32(-1);
for (size_t i = 0; i < dsize.height; i += BLOCK_SIZE)
{
size_t blockHeight = std::min<size_t>(BLOCK_SIZE, dsize.height - i);
for (size_t j = 0; j < dsize.width; j += BLOCK_SIZE)
{
size_t blockWidth = std::min<size_t>(BLOCK_SIZE, dsize.width - j);
// compute table
for (size_t y = 0; y < blockHeight; ++y)
{
s32 * map_row = getRowPtr(map, blockWidth * sizeof(s32) * 4, y);
f32 * coeff_row = getRowPtr(coeffs, blockWidth * sizeof(f32) * 2, y);
size_t x = 0, y_ = y + i;
f32 indeces[4] = { j + 0.0f, j + 1.0f, j + 2.0f, j + 3.0f };
float32x4_t v_x = vld1q_f32(indeces), v_y = vdupq_n_f32(y_), v_4 = vdupq_n_f32(4.0f);
float32x4_t v_yx = vmlaq_f32(v_m4, v_m2, v_y), v_yy = vmlaq_f32(v_m5, v_m3, v_y);
for ( ; x + 4 <= blockWidth; x += 4)
{
float32x4_t v_src_xf = vmlaq_f32(v_yx, v_m0, v_x);
float32x4_t v_src_yf = vmlaq_f32(v_yy, v_m1, v_x);
int32x4_t v_src_x0 = vcvtq_s32_f32(v_src_xf);
int32x4_t v_src_y0 = vcvtq_s32_f32(v_src_yf);
float32x4x2_t v_coeff;
v_coeff.val[0] = vsubq_f32(v_src_xf, vcvtq_f32_s32(v_src_x0));
v_coeff.val[1] = vsubq_f32(v_src_yf, vcvtq_f32_s32(v_src_y0));
uint32x4_t v_maskx = vcltq_f32(v_coeff.val[0], v_zero4f);
uint32x4_t v_masky = vcltq_f32(v_coeff.val[1], v_zero4f);
v_coeff.val[0] = vbslq_f32(v_maskx, vaddq_f32(v_one4f, v_coeff.val[0]), v_coeff.val[0]);
v_coeff.val[1] = vbslq_f32(v_masky, vaddq_f32(v_one4f, v_coeff.val[1]), v_coeff.val[1]);
v_src_x0 = vbslq_s32(v_maskx, vsubq_s32(v_src_x0, v_1), v_src_x0);
v_src_y0 = vbslq_s32(v_masky, vsubq_s32(v_src_y0, v_1), v_src_y0);
int32x4_t v_src_x1 = vaddq_s32(v_src_x0, v_1);
int32x4_t v_src_y1 = vaddq_s32(v_src_y0, v_1);
int32x4x4_t v_dst_index;
v_dst_index.val[0] = vmlaq_s32(v_src_x0, v_src_y0, v_step4);
v_dst_index.val[1] = vmlaq_s32(v_src_x1, v_src_y0, v_step4);
v_dst_index.val[2] = vmlaq_s32(v_src_x0, v_src_y1, v_step4);
v_dst_index.val[3] = vmlaq_s32(v_src_x1, v_src_y1, v_step4);
uint32x4_t v_mask_x0 = vandq_u32(vcgeq_f32(v_src_xf, v_zero4), vcleq_s32(v_src_x0, v_width4));
uint32x4_t v_mask_x1 = vandq_u32(vcgeq_f32(vaddq_f32(v_src_xf, v_one4f), v_zero4), vcleq_s32(v_src_x1, v_width4));
uint32x4_t v_mask_y0 = vandq_u32(vcgeq_f32(v_src_yf, v_zero4), vcleq_s32(v_src_y0, v_height4));
uint32x4_t v_mask_y1 = vandq_u32(vcgeq_f32(vaddq_f32(v_src_yf, v_one4f), v_zero4), vcleq_s32(v_src_y1, v_height4));
v_dst_index.val[0] = vbslq_s32(vandq_u32(v_mask_x0, v_mask_y0), v_dst_index.val[0], v_m1_4);
v_dst_index.val[1] = vbslq_s32(vandq_u32(v_mask_x1, v_mask_y0), v_dst_index.val[1], v_m1_4);
v_dst_index.val[2] = vbslq_s32(vandq_u32(v_mask_x0, v_mask_y1), v_dst_index.val[2], v_m1_4);
v_dst_index.val[3] = vbslq_s32(vandq_u32(v_mask_x1, v_mask_y1), v_dst_index.val[3], v_m1_4);
vst2q_f32(coeff_row + (x << 1), v_coeff);
vst4q_s32(map_row + (x << 2), v_dst_index);
v_x = vaddq_f32(v_x, v_4);
}
f32 yx = m[2] * y_ + m[4], yy = m[3] * y_ + m[5];
for (ptrdiff_t x_ = x + j; x < blockWidth; ++x, ++x_)
{
f32 src_x_f = m[0] * x_ + yx;
f32 src_y_f = m[1] * x_ + yy;
s32 src0_x = (s32)floorf(src_x_f), src1_x = src0_x + 1;
s32 src0_y = (s32)floorf(src_y_f), src1_y = src0_y + 1;
coeff_row[(x << 1) + 0] = src_x_f - src0_x;
coeff_row[(x << 1) + 1] = src_y_f - src0_y;
map_row[(x << 2) + 0] = (src0_x >= 0) && (src0_x < (s32)ssize.width) &&
(src0_y >= 0) && (src0_y < (s32)ssize.height) ? src0_y * srcStride + src0_x : -1;
map_row[(x << 2) + 1] = (src1_x >= 0) && (src1_x < (s32)ssize.width) &&
(src0_y >= 0) && (src0_y < (s32)ssize.height) ? src0_y * srcStride + src1_x : -1;
map_row[(x << 2) + 2] = (src0_x >= 0) && (src0_x < (s32)ssize.width) &&
(src1_y >= 0) && (src1_y < (s32)ssize.height) ? src1_y * srcStride + src0_x : -1;
map_row[(x << 2) + 3] = (src1_x >= 0) && (src1_x < (s32)ssize.width) &&
(src1_y >= 0) && (src1_y < (s32)ssize.height) ? src1_y * srcStride + src1_x : -1;
}
}
remapLinearConst(Size2D(blockWidth, blockHeight),
srcBase, &map[0], &coeffs[0],
getRowPtr(dstBase, dstStride, i) + j, dstStride, borderValue);
}
}
}
#else
(void)ssize;
(void)dsize;
(void)srcBase;
(void)srcStride;
(void)m;
(void)dstBase;
(void)dstStride;
(void)borderMode;
(void)borderValue;
#endif
}
} // namespace CAROTENE_NS