3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Sum Of Absolute Difference related code - |
* - Sum Of Absolute Difference related code - |
5 |
* |
* |
6 |
* Copyright(C) 2001-2003 Peter Ross <pross@xvid.org> |
* Copyright(C) 2001-2010 Peter Ross <pross@xvid.org> |
7 |
|
* 2010 Michael Militzer <michael@xvid.org> |
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 |
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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
22 |
* |
* |
23 |
* $Id: sad.c,v 1.16 2004-04-12 15:49:56 edgomez Exp $ |
* $Id: sad.c,v 1.17 2010-11-28 15:18:21 Isibaar Exp $ |
24 |
* |
* |
25 |
****************************************************************************/ |
****************************************************************************/ |
26 |
|
|
39 |
sse8Func_16bitPtr sse8_16bit; |
sse8Func_16bitPtr sse8_16bit; |
40 |
sse8Func_8bitPtr sse8_8bit; |
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 |
|
|
48 |
|
|
334 |
|
|
335 |
return(sse); |
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 |
|
} |