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
// Copyright (c) 2012-2017 VideoStitch SAS
// Copyright (c) 2018 stitchEm
#define BLOCK_SIZE 16
// ------------------- Subsampling
/**
* Subsample src by a factor of two on each dimension and put it in into dst.
* src has size (srcWidth * srcHeight), dst has size ((srcWidth + 1)/2 * (srcHeight + 1)/2)
*/
__global__ void subsample22RegularKernel(global_mem unsigned char* __restrict__ dst,
global_mem const unsigned char* __restrict__ src, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth, unsigned dstHeight) {
unsigned dstX = get_global_id_x();
unsigned dstY = get_global_id_y();
unsigned srcX = dstX * 2;
unsigned srcY = dstY * 2;
if (dstX < dstWidth && dstY < dstHeight && srcX + 1 < srcWidth && srcY + 1 < srcHeight) {
float v = src[srcY * srcWidth + srcX];
v += src[srcY * srcWidth + srcX + 1];
v += src[(srcY + 1) * srcWidth + srcX];
v += src[(srcY + 1) * srcWidth + srcX + 1];
dst[dstY * dstWidth + dstX] = (unsigned char)(v / 4.0f);
}
}
__global__ void subsample22RegularKernelFloat2(global_mem float2* __restrict__ dst,
global_mem const float2* __restrict__ src, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth, unsigned dstHeight) {
unsigned dstX = get_global_id_x();
unsigned dstY = get_global_id_y();
unsigned srcX = dstX * 2;
unsigned srcY = dstY * 2;
if (dstX < dstWidth && dstY < dstHeight && srcX + 1 < srcWidth && srcY + 1 < srcHeight) {
float2 v = src[srcY * srcWidth + srcX];
v += src[srcY * srcWidth + srcX + 1];
v += src[(srcY + 1) * srcWidth + srcX];
v += src[(srcY + 1) * srcWidth + srcX + 1];
dst[dstY * dstWidth + dstX] = v / 4.0f;
}
}
/**
* The pixels for this kernel have a bottom neighbour.
*/
__global__ void subsample22RightBoundaryKernel(global_mem unsigned char* __restrict__ dst,
global_mem const unsigned char* __restrict__ src, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth, unsigned dstHeight) {
unsigned dstX = (srcWidth - 1) / 2;
unsigned dstY = get_global_id_y();
unsigned srcX = srcWidth - 1;
unsigned srcY = dstY * 2;
if (dstX < dstWidth && dstY < dstHeight && srcY + 1 < srcHeight) {
float v = src[srcY * srcWidth + srcX];
v += src[(srcY + 1) * srcWidth + srcX];
dst[dstY * dstWidth + dstX] = (unsigned char)(v / 2.0f);
}
}
__global__ void subsample22RightBoundaryKernelFloat2(global_mem float2* __restrict__ dst,
global_mem const float2* __restrict__ src, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth, unsigned dstHeight) {
unsigned dstX = (srcWidth - 1) / 2;
unsigned dstY = get_global_id_y();
unsigned srcX = srcWidth - 1;
unsigned srcY = dstY * 2;
if (dstX < dstWidth && dstY < dstHeight && srcY + 1 < srcHeight) {
float2 v = src[srcY * srcWidth + srcX];
v += src[(srcY + 1) * srcWidth + srcX];
dst[dstY * dstWidth + dstX] = v / 2.0f;
}
}
/**
* The pixels for this kernel have a right neighbour
*/
__global__ void subsample22BottomBoundaryKernel(global_mem unsigned char* __restrict__ dst,
global_mem const unsigned char* __restrict__ src, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth, unsigned dstHeight) {
unsigned dstX = get_global_id_x();
unsigned dstY = (srcHeight - 1) / 2;
unsigned srcX = dstX * 2;
unsigned srcY = srcHeight - 1;
if (dstX < dstWidth && dstY < dstHeight && srcX + 1 < srcWidth) {
float v = src[srcY * srcWidth + srcX];
v += src[srcY * srcWidth + srcX + 1];
dst[dstY * dstWidth + dstX] = (unsigned char)(v / 2.0f);
}
}
__global__ void subsample22BottomBoundaryKernelFloat2(global_mem float2* __restrict__ dst,
global_mem const float2* __restrict__ src, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth, unsigned dstHeight) {
unsigned dstX = get_global_id_x();
unsigned dstY = (srcHeight - 1) / 2;
unsigned srcX = dstX * 2;
unsigned srcY = srcHeight - 1;
if (dstX < dstWidth && dstY < dstHeight && srcX + 1 < srcWidth) {
float2 v = src[srcY * srcWidth + srcX];
v += src[srcY * srcWidth + srcX + 1];
dst[dstY * dstWidth + dstX] = v / 2.0f;
}
}
__global__ void subsample22MaskRegularKernel(global_mem float2* __restrict__ dst,
global_mem uint32_t* __restrict__ dstMask,
global_mem const float2* __restrict__ src,
global_mem const uint32_t* __restrict__ srcMask, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth, unsigned dstHeight) {
unsigned dstX = get_global_id_x();
unsigned dstY = get_global_id_y();
unsigned srcX = dstX * 2;
unsigned srcY = dstY * 2;
if (dstX < dstWidth && dstY < dstHeight && srcX + 1 < srcWidth && srcY + 1 < srcHeight) {
// Here, weight is used as a mask
uint32_t w00 = srcMask[srcY * srcWidth + srcX];
uint32_t w01 = srcMask[srcY * srcWidth + srcX + 1];
uint32_t w10 = srcMask[(srcY + 1) * srcWidth + srcX];
uint32_t w11 = srcMask[(srcY + 1) * srcWidth + srcX + 1];
if (w00 == w01 && w01 == w10 && w10 == w11) {
float2 v = src[srcY * srcWidth + srcX];
v += src[srcY * srcWidth + srcX + 1];
v += src[(srcY + 1) * srcWidth + srcX];
v += src[(srcY + 1) * srcWidth + srcX + 1];
dst[dstY * dstWidth + dstX] = v / 4.0f;
} else {
dst[dstY * dstWidth + dstX] = src[srcY * srcWidth + srcX];
}
dstMask[dstY * dstWidth + dstX] = w00;
}
}
/**
* The pixels for this kernel have a bottom neighbour.
*/
__global__ void subsample22MaskRightBoundaryKernel(global_mem float2* __restrict__ dst,
global_mem uint32_t* __restrict__ dstMask,
global_mem const float2* __restrict__ src,
global_mem const uint32_t* __restrict__ srcMask, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth, unsigned dstHeight) {
unsigned dstX = (srcWidth - 1) / 2;
unsigned dstY = get_global_id_y();
unsigned srcX = srcWidth - 1;
unsigned srcY = dstY * 2;
if (dstX < dstWidth && dstY < dstHeight && srcY + 1 < srcHeight) {
uint32_t w00 = srcMask[srcY * srcWidth + srcX];
uint32_t w10 = srcMask[(srcY + 1) * srcWidth + srcX];
if (w00 == w10) {
float2 v = src[srcY * srcWidth + srcX];
v += src[(srcY + 1) * srcWidth + srcX];
dst[dstY * dstWidth + dstX] = v / 2.0f;
} else {
dst[dstY * dstWidth + dstX] = src[srcY * srcWidth + srcX];
}
dstMask[dstY * dstWidth + dstX] = w00;
}
}
/**
* The pixels for this kernel have a right neighbour
*/
__global__ void subsample22MaskBottomBoundaryKernel(global_mem float2* __restrict__ dst,
global_mem uint32_t* __restrict__ dstMask,
global_mem const float2* __restrict__ src,
global_mem const uint32_t* __restrict__ srcMask, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth, unsigned dstHeight) {
unsigned dstX = get_global_id_x();
unsigned dstY = (srcHeight - 1) / 2;
unsigned srcX = dstX * 2;
unsigned srcY = srcHeight - 1;
if (dstX < dstWidth && dstY < dstHeight && srcX + 1 < srcWidth) {
uint32_t w00 = srcMask[srcY * srcWidth + srcX];
uint32_t w01 = srcMask[srcY * srcWidth + srcX + 1];
if (w00 == w01) {
float2 v = src[srcY * srcWidth + srcX];
v += src[srcY * srcWidth + srcX + 1];
dst[dstY * dstWidth + dstX] = v / 2.0f;
} else {
dst[dstY * dstWidth + dstX] = src[srcY * srcWidth + srcX];
}
dstMask[dstY * dstWidth + dstX] = w00;
}
}
/**
* Subsample src by a factor of two on each dimension and put it in into dst.
* src has size (srcWidth * srcHeight), dst has size ((srcWidth + 1)/2 * (srcHeight + 1)/2)
*/
__global__ void subsample22NearestRegularKernel(global_mem unsigned char* __restrict__ dst,
global_mem const unsigned char* __restrict__ src, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth, unsigned dstHeight) {
unsigned dstX = get_global_id_x();
unsigned dstY = get_global_id_y();
unsigned srcX = dstX * 2;
unsigned srcY = dstY * 2;
if (dstX < dstWidth && dstY < dstHeight && srcX + 1 < srcWidth && srcY + 1 < srcHeight) {
dst[dstY * dstWidth + dstX] = src[srcY * srcWidth + srcX];
}
}
/**
* The pixels for this kernel have a bottom neighbour.
*/
__global__ void subsample22NearestRightBoundaryKernel(global_mem unsigned char* __restrict__ dst,
global_mem const unsigned char* __restrict__ src,
unsigned srcWidth, unsigned srcHeight, unsigned dstWidth,
unsigned dstHeight) {
unsigned dstX = (srcWidth - 1) / 2;
unsigned dstY = get_global_id_y();
unsigned srcX = srcWidth - 1;
unsigned srcY = dstY * 2;
if (dstX < dstWidth && dstY < dstHeight && srcY + 1 < srcHeight) {
dst[dstY * dstWidth + dstX] = src[srcY * srcWidth + srcX];
}
}
/**
* The pixels for this kernel have a right neighbour
*/
__global__ void subsample22NearestBottomBoundaryKernel(global_mem unsigned char* __restrict__ dst,
global_mem const unsigned char* __restrict__ src,
unsigned srcWidth, unsigned srcHeight, unsigned dstWidth,
unsigned dstHeight) {
unsigned dstX = get_global_id_x();
unsigned dstY = (srcHeight - 1) / 2;
unsigned srcX = dstX * 2;
unsigned srcY = srcHeight - 1;
if (dstX < dstWidth && dstY < dstHeight && srcX + 1 < srcWidth) {
dst[dstY * dstWidth + dstX] = src[srcY * srcWidth + srcX];
}
}
/**
* Same except that the alpha of the upsampled pixel is zero only if the alpha of all samples is 0.
* There is an interior kernel that handles pixels with neighbours on the left, right and diagonal,
* and a border kernel for each of the 3 remaining cases.
*/
__global__ void subsample22RGBARegularKernel(global_mem uint32_t* __restrict__ dst,
global_mem const uint32_t* __restrict__ src, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth) {
unsigned dstX = get_global_id_x();
unsigned dstY = get_global_id_y();
unsigned srcX = dstX * 2;
unsigned srcY = dstY * 2;
if (srcX + 1 < srcWidth && srcY + 1 < srcHeight) {
int32_t a = 0, r = 0, g = 0, b = 0;
uint32_t v = src[srcY * srcWidth + srcX];
if (Image_RGBA_a(v)) {
++a;
r += Image_RGBA_r(v);
g += Image_RGBA_g(v);
b += Image_RGBA_b(v);
}
v = src[srcY * srcWidth + srcX + 1];
if (Image_RGBA_a(v)) {
++a;
r += Image_RGBA_r(v);
g += Image_RGBA_g(v);
b += Image_RGBA_b(v);
}
v = src[(srcY + 1) * srcWidth + srcX];
if (Image_RGBA_a(v)) {
++a;
r += Image_RGBA_r(v);
g += Image_RGBA_g(v);
b += Image_RGBA_b(v);
}
v = src[(srcY + 1) * srcWidth + srcX + 1];
if (Image_RGBA_a(v)) {
++a;
r += Image_RGBA_r(v);
g += Image_RGBA_g(v);
b += Image_RGBA_b(v);
}
if (a) {
dst[dstY * dstWidth + dstX] = Image_RGBA_pack(r / a, g / a, b / a, 0xff);
} else {
dst[dstY * dstWidth + dstX] = 0;
}
}
}
/**
* The pixels for this kernel have a bottom neighbour.
*/
__global__ void subsample22RGBARightBoundaryKernel(global_mem uint32_t* __restrict__ dst,
global_mem const uint32_t* __restrict__ src, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth) {
unsigned dstX = get_global_id_x() + srcWidth / 2;
unsigned dstY = get_global_id_y();
unsigned srcX = dstX * 2;
unsigned srcY = dstY * 2;
if (srcY + 1 < srcHeight) {
int32_t a = 0, r = 0, g = 0, b = 0;
uint32_t v = src[srcY * srcWidth + srcX];
if (Image_RGBA_a(v)) {
++a;
r += Image_RGBA_r(v);
g += Image_RGBA_g(v);
b += Image_RGBA_b(v);
}
v = src[(srcY + 1) * srcWidth + srcX];
if (Image_RGBA_a(v)) {
++a;
r += Image_RGBA_r(v);
g += Image_RGBA_g(v);
b += Image_RGBA_b(v);
}
if (a) {
dst[dstY * dstWidth + dstX] = Image_RGBA_pack(r / a, g / a, b / a, 0xff);
} else {
dst[dstY * dstWidth + dstX] = 0;
}
}
}
/**
* The pixels for this kernel have a right neighbour
*/
__global__ void subsample22RGBABottomBoundaryKernel(global_mem uint32_t* __restrict__ dst,
global_mem const uint32_t* __restrict__ src, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth) {
unsigned dstX = get_global_id_x();
unsigned dstY = get_global_id_y() + srcHeight / 2;
unsigned srcX = dstX * 2;
unsigned srcY = dstY * 2;
if (srcX + 1 < srcWidth) {
int32_t a = 0, r = 0, g = 0, b = 0;
uint32_t v = src[srcY * srcWidth + srcX];
if (Image_RGBA_a(v)) {
++a;
r += Image_RGBA_r(v);
g += Image_RGBA_g(v);
b += Image_RGBA_b(v);
}
v = src[srcY * srcWidth + srcX + 1];
if (Image_RGBA_a(v)) {
++a;
r += Image_RGBA_r(v);
g += Image_RGBA_g(v);
b += Image_RGBA_b(v);
}
if (a) {
dst[dstY * dstWidth + dstX] = Image_RGBA_pack(r / a, g / a, b / a, 0xff);
} else {
dst[dstY * dstWidth + dstX] = 0;
}
}
}
// ---------------- Masks sampling
__global__ void subsampleMask22Kernel(global_mem unsigned char* __restrict__ dst,
global_mem const unsigned char* __restrict__ src, unsigned srcWidth,
unsigned srcHeight, unsigned dstWidth) {
unsigned dstX = get_global_id_x();
unsigned dstY = get_global_id_y();
unsigned srcX = dstX * 2;
unsigned srcY = dstY * 2;
if (srcX < srcWidth && srcY < srcHeight) {
unsigned char result = 0;
if (src[srcY * srcWidth + srcX] == 1) {
result = 1;
}
if (srcX + 1 < srcWidth && srcY < srcHeight && src[srcY * srcWidth + (srcX + 1)] == 1) {
result = 1;
}
if (srcX < srcWidth && srcY + 1 < srcHeight && src[(srcY + 1) * srcWidth + srcX] == 1) {
result = 1;
}
if (srcX + 1 < srcWidth && srcY + 1 < srcHeight && src[(srcY + 1) * srcWidth + (srcX + 1)] == 1) {
result = 1;
}
dst[dstY * dstWidth + dstX] = result;
}
}