1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* sum of absolute difference |
* - Sum Of Absolute Difference related code - |
5 |
* |
* |
6 |
* This program is an implementation of a part of one or more MPEG-4 |
* Copyright(C) 2001-2010 Peter Ross <pross@xvid.org> |
7 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* 2010 Michael Militzer <michael@xvid.org> |
|
* to use this software module in hardware or software products are |
|
|
* advised that its use may infringe existing patents or copyrights, and |
|
|
* any such use would be at such party's own risk. The original |
|
|
* developer of this software module and his/her company, and subsequent |
|
|
* editors and their companies, will have no liability for use of this |
|
|
* software or modifications or derivatives thereof. |
|
8 |
* |
* |
9 |
* This program is free software; you can redistribute it and/or modify |
* This program is free software; you can redistribute it and/or modify |
10 |
* it under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
18 |
* |
* |
19 |
* You should have received a copy of the GNU General Public License |
* You should have received a copy of the GNU General Public License |
20 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
21 |
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
22 |
* |
* |
23 |
*************************************************************************/ |
* $Id$ |
|
|
|
|
/************************************************************************** |
|
|
* |
|
|
* History: |
|
24 |
* |
* |
25 |
* 14.02.2002 added sad16bi_c() |
****************************************************************************/ |
|
* 10.11.2001 initial version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
|
* |
|
|
*************************************************************************/ |
|
|
|
|
26 |
|
|
27 |
#include "../portab.h" |
#include "../portab.h" |
28 |
#include "../global.h" |
#include "../global.h" |
36 |
sad8biFuncPtr sad8bi; /* not really sad16, but no difference in prototype */ |
sad8biFuncPtr sad8bi; /* not really sad16, but no difference in prototype */ |
37 |
dev16FuncPtr dev16; |
dev16FuncPtr dev16; |
38 |
sad16vFuncPtr sad16v; |
sad16vFuncPtr sad16v; |
39 |
|
sse8Func_16bitPtr sse8_16bit; |
40 |
|
sse8Func_8bitPtr sse8_8bit; |
41 |
|
|
42 |
|
sseh8Func_16bitPtr sseh8_16bit; |
43 |
|
coeff8_energyFunc_Ptr coeff8_energy; |
44 |
|
blocksum8Func_Ptr blocksum8; |
45 |
|
|
46 |
sadInitFuncPtr sadInit; |
sadInitFuncPtr sadInit; |
47 |
|
|
235 |
int32_t *sad) |
int32_t *sad) |
236 |
{ |
{ |
237 |
sad[0] = sad16(cur, ref, stride, 256*4096); |
sad[0] = sad16(cur, ref, stride, 256*4096); |
238 |
sad[1] = sad16(cur + 8, ref + 8, stride, 256*4096); |
sad[1] = sad16(cur + 16, ref + 16, stride, 256*4096); |
239 |
sad[2] = sad16(cur + 8*stride, ref + 8*stride, stride, 256*4096); |
sad[2] = sad16(cur + 16*stride, ref + 16*stride, stride, 256*4096); |
240 |
sad[3] = sad16(cur + 8*stride + 8, ref + 8*stride + 8, stride, 256*4096); |
sad[3] = sad16(cur + 16*stride + 16, ref + 16*stride + 16, stride, 256*4096); |
241 |
|
|
242 |
return sad[0]+sad[1]+sad[2]+sad[3]; |
return sad[0]+sad[1]+sad[2]+sad[3]; |
243 |
} |
} |
283 |
} |
} |
284 |
|
|
285 |
return MRSAD16_CORRFACTOR * sad; |
return MRSAD16_CORRFACTOR * sad; |
286 |
|
} |
287 |
|
|
288 |
|
uint32_t |
289 |
|
sse8_16bit_c(const int16_t * b1, |
290 |
|
const int16_t * b2, |
291 |
|
const uint32_t stride) |
292 |
|
{ |
293 |
|
int i; |
294 |
|
int sse = 0; |
295 |
|
|
296 |
|
for (i=0; i<8; i++) { |
297 |
|
sse += (b1[0] - b2[0])*(b1[0] - b2[0]); |
298 |
|
sse += (b1[1] - b2[1])*(b1[1] - b2[1]); |
299 |
|
sse += (b1[2] - b2[2])*(b1[2] - b2[2]); |
300 |
|
sse += (b1[3] - b2[3])*(b1[3] - b2[3]); |
301 |
|
sse += (b1[4] - b2[4])*(b1[4] - b2[4]); |
302 |
|
sse += (b1[5] - b2[5])*(b1[5] - b2[5]); |
303 |
|
sse += (b1[6] - b2[6])*(b1[6] - b2[6]); |
304 |
|
sse += (b1[7] - b2[7])*(b1[7] - b2[7]); |
305 |
|
|
306 |
|
b1 = (const int16_t*)((int8_t*)b1+stride); |
307 |
|
b2 = (const int16_t*)((int8_t*)b2+stride); |
308 |
|
} |
309 |
|
|
310 |
|
return(sse); |
311 |
|
} |
312 |
|
|
313 |
|
uint32_t |
314 |
|
sse8_8bit_c(const uint8_t * b1, |
315 |
|
const uint8_t * b2, |
316 |
|
const uint32_t stride) |
317 |
|
{ |
318 |
|
int i; |
319 |
|
int sse = 0; |
320 |
|
|
321 |
|
for (i=0; i<8; i++) { |
322 |
|
sse += (b1[0] - b2[0])*(b1[0] - b2[0]); |
323 |
|
sse += (b1[1] - b2[1])*(b1[1] - b2[1]); |
324 |
|
sse += (b1[2] - b2[2])*(b1[2] - b2[2]); |
325 |
|
sse += (b1[3] - b2[3])*(b1[3] - b2[3]); |
326 |
|
sse += (b1[4] - b2[4])*(b1[4] - b2[4]); |
327 |
|
sse += (b1[5] - b2[5])*(b1[5] - b2[5]); |
328 |
|
sse += (b1[6] - b2[6])*(b1[6] - b2[6]); |
329 |
|
sse += (b1[7] - b2[7])*(b1[7] - b2[7]); |
330 |
|
|
331 |
|
b1 = b1+stride; |
332 |
|
b2 = b2+stride; |
333 |
|
} |
334 |
|
|
335 |
|
return(sse); |
336 |
} |
} |
337 |
|
|
338 |
|
|
339 |
|
/* PSNR-HVS-M helper functions */ |
340 |
|
|
341 |
|
static const int16_t iMask_Coeff[64] = { |
342 |
|
0, 29788, 32767, 20479, 13653, 8192, 6425, 5372, |
343 |
|
27306, 27306, 23405, 17246, 12603, 5650, 5461, 5958, |
344 |
|
23405, 25205, 20479, 13653, 8192, 5749, 4749, 5851, |
345 |
|
23405, 19275, 14894, 11299, 6425, 3766, 4096, 5285, |
346 |
|
18204, 14894, 8856, 5851, 4819, 3006, 3181, 4255, |
347 |
|
13653, 9362, 5958, 5120, 4045, 3151, 2900, 3562, |
348 |
|
6687, 5120, 4201, 3766, 3181, 2708, 2730, 3244, |
349 |
|
4551, 3562, 3449, 3344, 2926, 3277, 3181, 3310 |
350 |
|
}; |
351 |
|
|
352 |
|
/* Calculate CSF weighted energy of DCT coefficients */ |
353 |
|
|
354 |
|
uint32_t |
355 |
|
coeff8_energy_c(const int16_t * dct) |
356 |
|
{ |
357 |
|
int x, y; |
358 |
|
uint32_t sum_a = 0; |
359 |
|
|
360 |
|
for (y = 0; y < 8; y += 2) { |
361 |
|
for (x = 0; x < 8; x += 2) { |
362 |
|
int16_t a0 = ((dct[y*8+x]<<4) * iMask_Coeff[y*8+x]) >> 16; |
363 |
|
int16_t a1 = ((dct[y*8+x+1]<<4) * iMask_Coeff[y*8+x+1]) >> 16; |
364 |
|
int16_t a2 = ((dct[(y+1)*8+x]<<4) * iMask_Coeff[(y+1)*8+x]) >> 16; |
365 |
|
int16_t a3 = ((dct[(y+1)*8+x+1]<<4) * iMask_Coeff[(y+1)*8+x+1]) >> 16; |
366 |
|
|
367 |
|
sum_a += ((a0*a0 + a1*a1 + a2*a2 + a3*a3) >> 3); |
368 |
|
} |
369 |
|
} |
370 |
|
|
371 |
|
return sum_a; |
372 |
|
} |
373 |
|
|
374 |
|
/* Calculate MSE of DCT coeffs reduced by masking effect */ |
375 |
|
|
376 |
|
uint32_t |
377 |
|
sseh8_16bit_c(const int16_t * cur, const int16_t * ref, uint16_t mask) |
378 |
|
{ |
379 |
|
int j, i; |
380 |
|
uint32_t mse_h = 0; |
381 |
|
|
382 |
|
for (j = 0; j < 8; j++) { |
383 |
|
for (i = 0; i < 8; i++) { |
384 |
|
uint32_t t = (mask * Inv_iMask_Coeff[j*8+i] + 32) >> 7; |
385 |
|
uint16_t u = abs(cur[j*8+i] - ref[j*8+i]) << 4; |
386 |
|
uint16_t thresh = (t < 65536) ? t : 65535; |
387 |
|
|
388 |
|
if (u < thresh) |
389 |
|
u = 0; /* The error is not perceivable */ |
390 |
|
else |
391 |
|
u -= thresh; |
392 |
|
|
393 |
|
u = ((u + iCSF_Round[j*8 + i]) * iCSF_Coeff[j*8 + i]) >> 16; |
394 |
|
|
395 |
|
mse_h += (uint32_t) (u * u); |
396 |
|
} |
397 |
|
} |
398 |
|
|
399 |
|
return mse_h; |
400 |
|
} |
401 |
|
|
402 |
|
/* Sums all pixels of 8x8 block */ |
403 |
|
|
404 |
|
uint32_t |
405 |
|
blocksum8_c(const int8_t * cur, int stride, uint16_t sums[4], uint32_t squares[4]) |
406 |
|
{ |
407 |
|
int i, j; |
408 |
|
uint32_t sum = 0; |
409 |
|
|
410 |
|
sums[0] = sums[1] = sums[2] = sums[3] = 0; |
411 |
|
squares[0] = squares[1] = squares[2] = squares[3] = 0; |
412 |
|
|
413 |
|
for (j = 0; j < 8; j++) { |
414 |
|
for (i = 0; i < 8; i++) { |
415 |
|
uint8_t p = cur[j*stride + i]; |
416 |
|
|
417 |
|
sums[(j>>2)*2 + (i>>2)] += p; |
418 |
|
squares[(j>>2)*2 + (i>>2)] += p*p; |
419 |
|
|
420 |
|
sum += p; |
421 |
|
} |
422 |
|
} |
423 |
|
|
424 |
|
return sum; |
425 |
|
} |