3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* image stuff |
* image stuff |
5 |
* |
* |
6 |
|
* This program is an implementation of a part of one or more MPEG-4 |
7 |
|
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
8 |
|
* to use this software module in hardware or software products are |
9 |
|
* advised that its use may infringe existing patents or copyrights, and |
10 |
|
* any such use would be at such party's own risk. The original |
11 |
|
* developer of this software module and his/her company, and subsequent |
12 |
|
* editors and their companies, will have no liability for use of this |
13 |
|
* software or modifications or derivatives thereof. |
14 |
|
* |
15 |
* This program is free software; you can redistribute it and/or modify |
* This program is free software; you can redistribute it and/or modify |
16 |
* 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 |
17 |
* the Free Software Foundation; either version 2 of the License, or |
* the Free Software Foundation; either version 2 of the License, or |
32 |
* |
* |
33 |
* History: |
* History: |
34 |
* |
* |
35 |
|
* 05.10.2002 support for interpolated images in qpel mode - Isibaar |
36 |
* 01.05.2002 BFRAME image-based u,v interpolation |
* 01.05.2002 BFRAME image-based u,v interpolation |
37 |
* 22.04.2002 added some B-frame support |
* 22.04.2002 added some B-frame support |
38 |
* 14.04.2002 added image_dump_yuvpgm(), added image_mad() |
* 14.04.2002 added image_dump_yuvpgm(), added image_mad() |
39 |
* XVID_CSP_USER input support |
* XVID_CSP_USER input support |
40 |
* 09.04.2002 PSNR calculations |
* 09.04.2002 PSNR calculations - Isibaar |
41 |
* 06.04.2002 removed interlaced edging from U,V blocks (as per spec) |
* 06.04.2002 removed interlaced edging from U,V blocks (as per spec) |
42 |
* 26.03.2002 interlacing support (field-based edging in set_edges) |
* 26.03.2002 interlacing support (field-based edging in set_edges) |
43 |
* 26.01.2002 rgb555, rgb565 |
* 26.01.2002 rgb555, rgb565 |
54 |
#include <math.h> |
#include <math.h> |
55 |
|
|
56 |
#include "../portab.h" |
#include "../portab.h" |
57 |
|
#include "../global.h" // XVID_CSP_XXX's |
58 |
#include "../xvid.h" // XVID_CSP_XXX's |
#include "../xvid.h" // XVID_CSP_XXX's |
59 |
#include "image.h" |
#include "image.h" |
60 |
#include "colorspace.h" |
#include "colorspace.h" |
61 |
#include "interpolate8x8.h" |
#include "interpolate8x8.h" |
62 |
|
#include "reduced.h" |
63 |
#include "../divx4.h" |
#include "../divx4.h" |
64 |
#include "../utils/mem_align.h" |
#include "../utils/mem_align.h" |
65 |
|
|
66 |
|
#include "font.h" // XXX: remove later |
67 |
|
|
68 |
#define SAFETY 64 |
#define SAFETY 64 |
69 |
#define EDGE_SIZE2 (EDGE_SIZE/2) |
#define EDGE_SIZE2 (EDGE_SIZE/2) |
70 |
|
|
165 |
uint32_t edged_width, |
uint32_t edged_width, |
166 |
uint32_t edged_height, |
uint32_t edged_height, |
167 |
uint32_t width, |
uint32_t width, |
168 |
uint32_t height, |
uint32_t height) |
|
uint32_t interlacing) |
|
169 |
{ |
{ |
170 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
171 |
const uint32_t width2 = width / 2; |
const uint32_t width2 = width / 2; |
178 |
src = image->y; |
src = image->y; |
179 |
|
|
180 |
for (i = 0; i < EDGE_SIZE; i++) { |
for (i = 0; i < EDGE_SIZE; i++) { |
|
// if interlacing, edges contain top-most data from each field |
|
|
if (interlacing && (i & 1)) { |
|
|
memset(dst, *(src + edged_width), EDGE_SIZE); |
|
|
memcpy(dst + EDGE_SIZE, src + edged_width, width); |
|
|
memset(dst + edged_width - EDGE_SIZE, |
|
|
*(src + edged_width + width - 1), EDGE_SIZE); |
|
|
} else { |
|
181 |
memset(dst, *src, EDGE_SIZE); |
memset(dst, *src, EDGE_SIZE); |
182 |
memcpy(dst + EDGE_SIZE, src, width); |
memcpy(dst + EDGE_SIZE, src, width); |
183 |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
184 |
EDGE_SIZE); |
EDGE_SIZE); |
|
} |
|
185 |
dst += edged_width; |
dst += edged_width; |
186 |
} |
} |
187 |
|
|
194 |
|
|
195 |
src -= edged_width; |
src -= edged_width; |
196 |
for (i = 0; i < EDGE_SIZE; i++) { |
for (i = 0; i < EDGE_SIZE; i++) { |
|
// if interlacing, edges contain bottom-most data from each field |
|
|
if (interlacing && !(i & 1)) { |
|
|
memset(dst, *(src - edged_width), EDGE_SIZE); |
|
|
memcpy(dst + EDGE_SIZE, src - edged_width, width); |
|
|
memset(dst + edged_width - EDGE_SIZE, |
|
|
*(src - edged_width + width - 1), EDGE_SIZE); |
|
|
} else { |
|
197 |
memset(dst, *src, EDGE_SIZE); |
memset(dst, *src, EDGE_SIZE); |
198 |
memcpy(dst + EDGE_SIZE, src, width); |
memcpy(dst + EDGE_SIZE, src, width); |
199 |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
200 |
EDGE_SIZE); |
EDGE_SIZE); |
|
} |
|
201 |
dst += edged_width; |
dst += edged_width; |
202 |
} |
} |
203 |
|
|
266 |
IMAGE * refhv, |
IMAGE * refhv, |
267 |
uint32_t edged_width, |
uint32_t edged_width, |
268 |
uint32_t edged_height, |
uint32_t edged_height, |
269 |
|
uint32_t quarterpel, |
270 |
uint32_t rounding) |
uint32_t rounding) |
271 |
{ |
{ |
272 |
const uint32_t offset = EDGE_SIZE * (edged_width + 1); |
const uint32_t offset = EDGE_SIZE2 * (edged_width + 1); // we only interpolate half of the edge area |
273 |
const uint32_t stride_add = 7 * edged_width; |
const uint32_t stride_add = 7 * edged_width; |
274 |
|
/* |
275 |
#ifdef BFRAMES |
#ifdef BFRAMES |
276 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
277 |
const uint32_t edged_height2 = edged_height / 2; |
const uint32_t edged_height2 = edged_height / 2; |
278 |
const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); |
const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); |
279 |
const uint32_t stride_add2 = 7 * edged_width2; |
const uint32_t stride_add2 = 7 * edged_width2; |
280 |
#endif |
#endif |
281 |
|
*/ |
282 |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
283 |
uint32_t x, y; |
uint32_t x, y; |
284 |
|
|
293 |
v_ptr -= offset; |
v_ptr -= offset; |
294 |
hv_ptr -= offset; |
hv_ptr -= offset; |
295 |
|
|
296 |
for (y = 0; y < edged_height; y = y + 8) { |
if(quarterpel) { |
297 |
for (x = 0; x < edged_width; x = x + 8) { |
|
298 |
|
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
299 |
|
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
300 |
|
interpolate8x8_6tap_lowpass_h(h_ptr, n_ptr, edged_width, rounding); |
301 |
|
interpolate8x8_6tap_lowpass_v(v_ptr, n_ptr, edged_width, rounding); |
302 |
|
|
303 |
|
n_ptr += 8; |
304 |
|
h_ptr += 8; |
305 |
|
v_ptr += 8; |
306 |
|
} |
307 |
|
|
308 |
|
n_ptr += EDGE_SIZE; |
309 |
|
h_ptr += EDGE_SIZE; |
310 |
|
v_ptr += EDGE_SIZE; |
311 |
|
|
312 |
|
h_ptr += stride_add; |
313 |
|
v_ptr += stride_add; |
314 |
|
n_ptr += stride_add; |
315 |
|
} |
316 |
|
|
317 |
|
h_ptr = refh->y; |
318 |
|
h_ptr -= offset; |
319 |
|
|
320 |
|
for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { |
321 |
|
for (x = 0; x < (edged_width - EDGE_SIZE); x = x + 8) { |
322 |
|
interpolate8x8_6tap_lowpass_v(hv_ptr, h_ptr, edged_width, rounding); |
323 |
|
hv_ptr += 8; |
324 |
|
h_ptr += 8; |
325 |
|
} |
326 |
|
|
327 |
|
hv_ptr += EDGE_SIZE; |
328 |
|
h_ptr += EDGE_SIZE; |
329 |
|
|
330 |
|
hv_ptr += stride_add; |
331 |
|
h_ptr += stride_add; |
332 |
|
} |
333 |
|
} |
334 |
|
else { |
335 |
|
|
336 |
|
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
337 |
|
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
338 |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); |
339 |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); |
340 |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); |
344 |
v_ptr += 8; |
v_ptr += 8; |
345 |
hv_ptr += 8; |
hv_ptr += 8; |
346 |
} |
} |
347 |
|
|
348 |
|
h_ptr += EDGE_SIZE; |
349 |
|
v_ptr += EDGE_SIZE; |
350 |
|
hv_ptr += EDGE_SIZE; |
351 |
|
n_ptr += EDGE_SIZE; |
352 |
|
|
353 |
h_ptr += stride_add; |
h_ptr += stride_add; |
354 |
v_ptr += stride_add; |
v_ptr += stride_add; |
355 |
hv_ptr += stride_add; |
hv_ptr += stride_add; |
356 |
n_ptr += stride_add; |
n_ptr += stride_add; |
357 |
} |
} |
358 |
|
} |
359 |
|
/* |
360 |
#ifdef BFRAMES |
#ifdef BFRAMES |
361 |
n_ptr = refn->u; |
n_ptr = refn->u; |
362 |
h_ptr = refh->u; |
h_ptr = refh->u; |
368 |
v_ptr -= offset2; |
v_ptr -= offset2; |
369 |
hv_ptr -= offset2; |
hv_ptr -= offset2; |
370 |
|
|
371 |
for (y = 0; y < edged_height2; y = y + 8) { |
for (y = 0; y < edged_height2; y += 8) { |
372 |
for (x = 0; x < edged_width2; x = x + 8) { |
for (x = 0; x < edged_width2; x += 8) { |
373 |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width2, rounding); |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width2, rounding); |
374 |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width2, rounding); |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width2, rounding); |
375 |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width2, rounding); |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width2, rounding); |
412 |
n_ptr += stride_add2; |
n_ptr += stride_add2; |
413 |
} |
} |
414 |
#endif |
#endif |
415 |
|
*/ |
416 |
/* |
/* |
417 |
interpolate_halfpel_h( |
interpolate_halfpel_h( |
418 |
refh->y - offset, |
refh->y - offset, |
476 |
} |
} |
477 |
|
|
478 |
|
|
479 |
|
/* |
480 |
|
chroma optimize filter, invented by mf |
481 |
|
a chroma pixel is average from the surrounding pixels, when the |
482 |
|
correpsonding luma pixels are pure black or white. |
483 |
|
*/ |
484 |
|
|
485 |
|
void |
486 |
|
image_chroma_optimize(IMAGE * img, int width, int height, int edged_width) |
487 |
|
{ |
488 |
|
int x,y; |
489 |
|
int pixels = 0; |
490 |
|
|
491 |
|
for (y = 1; y < height/2 - 1; y++) |
492 |
|
for (x = 1; x < width/2 - 1; x++) |
493 |
|
{ |
494 |
|
#define IS_PURE(a) ((a)<=16||(a)>=235) |
495 |
|
#define IMG_Y(Y,X) img->y[(Y)*edged_width + (X)] |
496 |
|
#define IMG_U(Y,X) img->u[(Y)*edged_width/2 + (X)] |
497 |
|
#define IMG_V(Y,X) img->v[(Y)*edged_width/2 + (X)] |
498 |
|
|
499 |
|
if (IS_PURE(IMG_Y(y*2 ,x*2 )) && |
500 |
|
IS_PURE(IMG_Y(y*2 ,x*2+1)) && |
501 |
|
IS_PURE(IMG_Y(y*2+1,x*2 )) && |
502 |
|
IS_PURE(IMG_Y(y*2+1,x*2+1))) |
503 |
|
{ |
504 |
|
IMG_U(y,x) = (IMG_U(y,x-1) + IMG_U(y-1, x) + IMG_U(y, x+1) + IMG_U(y+1, x)) / 4; |
505 |
|
IMG_V(y,x) = (IMG_V(y,x-1) + IMG_V(y-1, x) + IMG_V(y, x+1) + IMG_V(y+1, x)) / 4; |
506 |
|
pixels++; |
507 |
|
} |
508 |
|
|
509 |
|
#undef IS_PURE |
510 |
|
#undef IMG_Y |
511 |
|
#undef IMG_U |
512 |
|
#undef IMG_V |
513 |
|
} |
514 |
|
|
515 |
|
DPRINTF(DPRINTF_DEBUG,"chroma_optimized_pixels = %i/%i", pixels, width*height/4); |
516 |
|
} |
517 |
|
|
518 |
|
|
519 |
|
|
520 |
|
|
521 |
|
|
522 |
|
/* |
523 |
|
perform safe packed colorspace conversion, by splitting |
524 |
|
the image up into an optimized area (pixel width divisible by 16), |
525 |
|
and two unoptimized/plain-c areas (pixel width divisible by 2) |
526 |
|
*/ |
527 |
|
|
528 |
|
static void |
529 |
|
safe_packed_conv(uint8_t * x_ptr, int x_stride, |
530 |
|
uint8_t * y_ptr, uint8_t * u_ptr, uint8_t * v_ptr, |
531 |
|
int y_stride, int uv_stride, |
532 |
|
int width, int height, int vflip, |
533 |
|
packedFunc * func_opt, packedFunc func_c, int size) |
534 |
|
{ |
535 |
|
int width_opt, width_c; |
536 |
|
|
537 |
|
if (func_opt != func_c && x_stride < size*((width+15)/16)*16) |
538 |
|
{ |
539 |
|
width_opt = width & (~15); |
540 |
|
width_c = width - width_opt; |
541 |
|
} |
542 |
|
else |
543 |
|
{ |
544 |
|
width_opt = width; |
545 |
|
width_c = 0; |
546 |
|
} |
547 |
|
|
548 |
|
func_opt(x_ptr, x_stride, |
549 |
|
y_ptr, u_ptr, v_ptr, y_stride, uv_stride, |
550 |
|
width_opt, height, vflip); |
551 |
|
|
552 |
|
if (width_c) |
553 |
|
{ |
554 |
|
func_c(x_ptr + size*width_opt, x_stride, |
555 |
|
y_ptr + width_opt, u_ptr + width_opt/2, v_ptr + width_opt/2, |
556 |
|
y_stride, uv_stride, width_c, height, vflip); |
557 |
|
} |
558 |
|
} |
559 |
|
|
560 |
|
|
561 |
|
|
562 |
int |
int |
563 |
image_input(IMAGE * image, |
image_input(IMAGE * image, |
564 |
uint32_t width, |
uint32_t width, |
565 |
int height, |
int height, |
566 |
uint32_t edged_width, |
uint32_t edged_width, |
567 |
uint8_t * src, |
uint8_t * src, |
568 |
int csp) |
int src_stride, |
569 |
|
int csp, |
570 |
|
int interlacing) |
571 |
{ |
{ |
572 |
|
const int edged_width2 = edged_width/2; |
573 |
|
const int width2 = width/2; |
574 |
|
const int height2 = height/2; |
575 |
|
//const int height_signed = (csp & XVID_CSP_VFLIP) ? -height : height; |
576 |
|
|
577 |
|
|
578 |
/* if (csp & XVID_CSP_VFLIP) |
// int src_stride = width; |
579 |
|
|
580 |
|
// --- xvid 2.1 compatiblity patch --- |
581 |
|
// --- remove when xvid_dec_frame->stride equals real stride |
582 |
|
/* |
583 |
|
if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB555 || |
584 |
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB565 || |
585 |
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_YUY2 || |
586 |
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_YVYU || |
587 |
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_UYVY) |
588 |
|
{ |
589 |
|
src_stride *= 2; |
590 |
|
} |
591 |
|
else if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB24) |
592 |
{ |
{ |
593 |
height = -height; |
src_stride *= 3; |
594 |
|
} |
595 |
|
else if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB32 || |
596 |
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_ABGR || |
597 |
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGBA) |
598 |
|
{ |
599 |
|
src_stride *= 4; |
600 |
} |
} |
601 |
*/ |
*/ |
602 |
|
// ^--- xvid 2.1 compatiblity fix ---^ |
603 |
|
|
604 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
605 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
606 |
rgb555_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
607 |
edged_width); |
src, src_stride, image->y, image->u, image->v, |
608 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
609 |
|
interlacing?rgb555i_to_yv12 :rgb555_to_yv12, |
610 |
|
interlacing?rgb555i_to_yv12_c:rgb555_to_yv12_c, 2); |
611 |
|
break; |
612 |
|
|
613 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
614 |
rgb565_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
615 |
edged_width); |
src, src_stride, image->y, image->u, image->v, |
616 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
617 |
|
interlacing?rgb565i_to_yv12 :rgb565_to_yv12, |
618 |
|
interlacing?rgb565i_to_yv12_c:rgb565_to_yv12_c, 2); |
619 |
|
break; |
620 |
|
|
621 |
|
|
622 |
case XVID_CSP_RGB24: |
case XVID_CSP_RGB24: |
623 |
rgb24_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
624 |
edged_width); |
src, src_stride, image->y, image->u, image->v, |
625 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
626 |
|
interlacing?bgri_to_yv12 :bgr_to_yv12, |
627 |
|
interlacing?bgri_to_yv12_c:bgr_to_yv12_c, 3); |
628 |
|
break; |
629 |
|
|
630 |
case XVID_CSP_RGB32: |
case XVID_CSP_RGB32: |
631 |
rgb32_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
632 |
edged_width); |
src, src_stride, image->y, image->u, image->v, |
633 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
634 |
|
interlacing?bgrai_to_yv12 :bgra_to_yv12, |
635 |
|
interlacing?bgrai_to_yv12_c:bgra_to_yv12_c, 4); |
636 |
|
break; |
637 |
|
|
638 |
case XVID_CSP_I420: |
case XVID_CSP_ABGR : |
639 |
yuv_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
640 |
edged_width); |
src, src_stride, image->y, image->u, image->v, |
641 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
642 |
|
interlacing?abgri_to_yv12 :abgr_to_yv12, |
643 |
|
interlacing?abgri_to_yv12_c:abgr_to_yv12_c, 4); |
644 |
|
break; |
645 |
|
|
646 |
case XVID_CSP_YV12: /* u/v swapped */ |
case XVID_CSP_RGBA : |
647 |
yuv_to_yv12(image->y, image->v, image->u, src, width, height, |
safe_packed_conv( |
648 |
edged_width); |
src, src_stride, image->y, image->u, image->v, |
649 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
650 |
|
interlacing?rgbai_to_yv12 :rgba_to_yv12, |
651 |
|
interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4); |
652 |
|
break; |
653 |
|
|
654 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
655 |
yuyv_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
656 |
edged_width); |
src, src_stride, image->y, image->u, image->v, |
657 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
658 |
|
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
659 |
|
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
660 |
|
break; |
661 |
|
|
662 |
case XVID_CSP_YVYU: /* u/v swapped */ |
case XVID_CSP_YVYU: /* u/v swapped */ |
663 |
yuyv_to_yv12(image->y, image->v, image->u, src, width, height, |
safe_packed_conv( |
664 |
edged_width); |
src, src_stride, image->y, image->v, image->y, |
665 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
666 |
|
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
667 |
|
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
668 |
|
break; |
669 |
|
|
670 |
case XVID_CSP_UYVY: |
case XVID_CSP_UYVY: |
671 |
uyvy_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
672 |
edged_width); |
src, src_stride, image->y, image->u, image->v, |
673 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
674 |
|
interlacing?uyvyi_to_yv12 :uyvy_to_yv12, |
675 |
|
interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2); |
676 |
|
break; |
677 |
|
|
678 |
|
case XVID_CSP_I420: |
679 |
|
yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
680 |
|
src, src + src_stride*height, src + src_stride*height + (src_stride/2)*height2, |
681 |
|
src_stride, src_stride/2, width, height, (csp & XVID_CSP_VFLIP)); |
682 |
|
break |
683 |
|
; |
684 |
|
case XVID_CSP_YV12: /* u/v swapped */ |
685 |
|
yv12_to_yv12(image->y, image->v, image->u, edged_width, edged_width2, |
686 |
|
src, src + src_stride*height, src + src_stride*height + (src_stride/2)*height2, |
687 |
|
src_stride, src_stride/2, width, height, (csp & XVID_CSP_VFLIP)); |
688 |
|
break; |
689 |
|
|
690 |
case XVID_CSP_USER: |
case XVID_CSP_USER: |
691 |
user_to_yuv_c(image->y, image->u, image->v, edged_width, |
{ |
692 |
(DEC_PICTURE *) src, width, height); |
DEC_PICTURE * pic = (DEC_PICTURE*)src; |
693 |
return 0; |
yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
694 |
|
pic->y, pic->u, pic->v, pic->stride_y, pic->stride_y, |
695 |
|
width, height, (csp & XVID_CSP_VFLIP)); |
696 |
|
} |
697 |
|
break; |
698 |
|
|
699 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
700 |
break; |
break; |
701 |
|
|
702 |
|
default : |
703 |
|
return -1; |
704 |
|
} |
705 |
|
|
706 |
|
|
707 |
|
/* pad out image when the width and/or height is not a multiple of 16 */ |
708 |
|
|
709 |
|
if (width & 15) |
710 |
|
{ |
711 |
|
int i; |
712 |
|
int pad_width = 16 - (width&15); |
713 |
|
for (i = 0; i < height; i++) |
714 |
|
{ |
715 |
|
memset(image->y + i*edged_width + width, |
716 |
|
*(image->y + i*edged_width + width - 1), pad_width); |
717 |
|
} |
718 |
|
for (i = 0; i < height/2; i++) |
719 |
|
{ |
720 |
|
memset(image->u + i*edged_width2 + width2, |
721 |
|
*(image->u + i*edged_width2 + width2 - 1),pad_width/2); |
722 |
|
memset(image->v + i*edged_width2 + width2, |
723 |
|
*(image->v + i*edged_width2 + width2 - 1),pad_width/2); |
724 |
|
} |
725 |
} |
} |
726 |
|
|
727 |
return -1; |
if (height & 15) |
728 |
|
{ |
729 |
|
int pad_height = 16 - (height&15); |
730 |
|
int length = ((width+15)/16)*16; |
731 |
|
int i; |
732 |
|
for (i = 0; i < pad_height; i++) |
733 |
|
{ |
734 |
|
memcpy(image->y + (height+i)*edged_width, |
735 |
|
image->y + (height-1)*edged_width,length); |
736 |
|
} |
737 |
|
|
738 |
|
for (i = 0; i < pad_height/2; i++) |
739 |
|
{ |
740 |
|
memcpy(image->u + (height2+i)*edged_width2, |
741 |
|
image->u + (height2-1)*edged_width2,length/2); |
742 |
|
memcpy(image->v + (height2+i)*edged_width2, |
743 |
|
image->v + (height2-1)*edged_width2,length/2); |
744 |
|
} |
745 |
|
} |
746 |
|
|
747 |
|
/* |
748 |
|
if (interlacing) |
749 |
|
image_printf(image, edged_width, height, 5,5, "[i]"); |
750 |
|
image_dump_yuvpgm(image, edged_width, ((width+15)/16)*16, ((height+15)/16)*16, "\\encode.pgm"); |
751 |
|
*/ |
752 |
|
return 0; |
753 |
} |
} |
754 |
|
|
755 |
|
|
761 |
uint32_t edged_width, |
uint32_t edged_width, |
762 |
uint8_t * dst, |
uint8_t * dst, |
763 |
uint32_t dst_stride, |
uint32_t dst_stride, |
764 |
int csp) |
int csp, |
765 |
|
int interlacing) |
766 |
|
{ |
767 |
|
const int edged_width2 = edged_width/2; |
768 |
|
int height2 = height/2; |
769 |
|
|
770 |
|
/* |
771 |
|
if (interlacing) |
772 |
|
image_printf(image, edged_width, height, 5,100, "[i]=%i,%i",width,height); |
773 |
|
image_dump_yuvpgm(image, edged_width, width, height, "\\decode.pgm"); |
774 |
|
*/ |
775 |
|
|
776 |
|
|
777 |
|
// --- xvid 2.1 compatiblity patch --- |
778 |
|
// --- remove when xvid_dec_frame->stride equals real stride |
779 |
|
/* |
780 |
|
if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB555 || |
781 |
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB565 || |
782 |
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_YUY2 || |
783 |
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_YVYU || |
784 |
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_UYVY) |
785 |
|
{ |
786 |
|
dst_stride *= 2; |
787 |
|
} |
788 |
|
else if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB24) |
789 |
|
{ |
790 |
|
dst_stride *= 3; |
791 |
|
} |
792 |
|
else if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB32 || |
793 |
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_ABGR || |
794 |
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGBA) |
795 |
{ |
{ |
796 |
if (csp & XVID_CSP_VFLIP) { |
dst_stride *= 4; |
|
height = -height; |
|
797 |
} |
} |
798 |
|
*/ |
799 |
|
// ^--- xvid 2.1 compatiblity fix ---^ |
800 |
|
|
801 |
|
|
802 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
803 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
804 |
yv12_to_rgb555(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
805 |
edged_width, edged_width / 2, width, height); |
dst, dst_stride, image->y, image->u, image->v, |
806 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
807 |
|
interlacing?yv12_to_rgb555i :yv12_to_rgb555, |
808 |
|
interlacing?yv12_to_rgb555i_c:yv12_to_rgb555_c, 2); |
809 |
return 0; |
return 0; |
810 |
|
|
811 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
812 |
yv12_to_rgb565(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
813 |
edged_width, edged_width / 2, width, height); |
dst, dst_stride, image->y, image->u, image->v, |
814 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
815 |
|
interlacing?yv12_to_rgb565i :yv12_to_rgb565, |
816 |
|
interlacing?yv12_to_rgb565i_c:yv12_to_rgb565_c, 2); |
817 |
return 0; |
return 0; |
818 |
|
|
819 |
case XVID_CSP_RGB24: |
case XVID_CSP_RGB24: |
820 |
yv12_to_rgb24(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
821 |
edged_width, edged_width / 2, width, height); |
dst, dst_stride, image->y, image->u, image->v, |
822 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
823 |
|
interlacing?yv12_to_bgri :yv12_to_bgr, |
824 |
|
interlacing?yv12_to_bgri_c:yv12_to_bgr_c, 3); |
825 |
return 0; |
return 0; |
826 |
|
|
827 |
case XVID_CSP_RGB32: |
case XVID_CSP_RGB32: |
828 |
yv12_to_rgb32(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
829 |
edged_width, edged_width / 2, width, height); |
dst, dst_stride, image->y, image->u, image->v, |
830 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
831 |
|
interlacing?yv12_to_bgrai :yv12_to_bgra, |
832 |
|
interlacing?yv12_to_bgrai_c:yv12_to_bgra_c, 4); |
833 |
|
return 0; |
834 |
|
|
835 |
|
case XVID_CSP_ABGR: |
836 |
|
safe_packed_conv( |
837 |
|
dst, dst_stride, image->y, image->u, image->v, |
838 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
839 |
|
interlacing?yv12_to_abgri :yv12_to_abgr, |
840 |
|
interlacing?yv12_to_abgri_c:yv12_to_abgr_c, 4); |
841 |
|
return 0; |
842 |
|
|
843 |
|
case XVID_CSP_RGBA: |
844 |
|
safe_packed_conv( |
845 |
|
dst, dst_stride, image->y, image->u, image->v, |
846 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
847 |
|
interlacing?yv12_to_rgbai :yv12_to_rgba, |
848 |
|
interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4); |
849 |
return 0; |
return 0; |
850 |
|
|
851 |
case XVID_CSP_I420: |
case XVID_CSP_YUY2: |
852 |
yv12_to_yuv(dst, dst_stride, image->y, image->u, image->v, edged_width, |
safe_packed_conv( |
853 |
edged_width / 2, width, height); |
dst, dst_stride, image->y, image->u, image->v, |
854 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
855 |
|
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
856 |
|
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
857 |
return 0; |
return 0; |
858 |
|
|
859 |
case XVID_CSP_YV12: // u,v swapped |
case XVID_CSP_YVYU: // u,v swapped |
860 |
yv12_to_yuv(dst, dst_stride, image->y, image->v, image->u, edged_width, |
safe_packed_conv( |
861 |
edged_width / 2, width, height); |
dst, dst_stride, image->y, image->v, image->u, |
862 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
863 |
|
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
864 |
|
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
865 |
return 0; |
return 0; |
866 |
|
|
867 |
case XVID_CSP_YUY2: |
case XVID_CSP_UYVY: |
868 |
yv12_to_yuyv(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
869 |
edged_width, edged_width / 2, width, height); |
dst, dst_stride, image->y, image->u, image->v, |
870 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
871 |
|
interlacing?yv12_to_uyvyi :yv12_to_uyvy, |
872 |
|
interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2); |
873 |
return 0; |
return 0; |
874 |
|
|
875 |
case XVID_CSP_YVYU: // u,v swapped |
case XVID_CSP_I420: |
876 |
yv12_to_yuyv(dst, dst_stride, image->y, image->v, image->u, |
yv12_to_yv12(dst, dst + dst_stride*height, dst + dst_stride*height + (dst_stride/2)*height2, |
877 |
edged_width, edged_width / 2, width, height); |
dst_stride, dst_stride/2, |
878 |
|
image->y, image->u, image->v, edged_width, edged_width2, |
879 |
|
width, height, (csp & XVID_CSP_VFLIP)); |
880 |
return 0; |
return 0; |
881 |
|
|
882 |
case XVID_CSP_UYVY: |
case XVID_CSP_YV12: // u,v swapped |
883 |
yv12_to_uyvy(dst, dst_stride, image->y, image->u, image->v, |
yv12_to_yv12(dst, dst + dst_stride*height, dst + dst_stride*height + (dst_stride/2)*height2, |
884 |
edged_width, edged_width / 2, width, height); |
dst_stride, dst_stride/2, |
885 |
|
image->y, image->v, image->u, edged_width, edged_width2, |
886 |
|
width, height, (csp & XVID_CSP_VFLIP)); |
887 |
return 0; |
return 0; |
888 |
|
|
889 |
case XVID_CSP_USER: |
case XVID_CSP_USER: |
890 |
((DEC_PICTURE *) dst)->y = image->y; |
{ |
891 |
((DEC_PICTURE *) dst)->u = image->u; |
DEC_PICTURE * pic = (DEC_PICTURE*)dst; |
892 |
((DEC_PICTURE *) dst)->v = image->v; |
pic->y = image->y; |
893 |
((DEC_PICTURE *) dst)->stride_y = edged_width; |
pic->u = image->u; |
894 |
((DEC_PICTURE *) dst)->stride_uv = edged_width / 2; |
pic->v = image->v; |
895 |
|
pic->stride_y = edged_width; |
896 |
|
pic->stride_uv = edged_width / 2; |
897 |
|
} |
898 |
return 0; |
return 0; |
899 |
|
|
900 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
938 |
return psnr_y; |
return psnr_y; |
939 |
} |
} |
940 |
|
|
941 |
|
|
942 |
|
float sse_to_PSNR(long sse, int pixels) |
943 |
|
{ |
944 |
|
if (sse==0) |
945 |
|
return 99.99F; |
946 |
|
|
947 |
|
return 48.131F - 10*(float)log10((float)sse/(float)(pixels)); // log10(255*255)=4.8131 |
948 |
|
|
949 |
|
} |
950 |
|
|
951 |
|
long plane_sse(uint8_t * orig, |
952 |
|
uint8_t * recon, |
953 |
|
uint16_t stride, |
954 |
|
uint16_t width, |
955 |
|
uint16_t height) |
956 |
|
{ |
957 |
|
int diff, x, y; |
958 |
|
long sse=0; |
959 |
|
|
960 |
|
for (y = 0; y < height; y++) { |
961 |
|
for (x = 0; x < width; x++) { |
962 |
|
diff = *(orig + x) - *(recon + x); |
963 |
|
sse += diff * diff; |
964 |
|
} |
965 |
|
orig += stride; |
966 |
|
recon += stride; |
967 |
|
} |
968 |
|
return sse; |
969 |
|
} |
970 |
|
|
971 |
/* |
/* |
972 |
|
|
973 |
#include <stdio.h> |
#include <stdio.h> |
1064 |
} |
} |
1065 |
|
|
1066 |
|
|
|
#define ABS(X) (((X)>0)?(X):-(X)) |
|
1067 |
float |
float |
1068 |
image_mad(const IMAGE * img1, |
image_mad(const IMAGE * img1, |
1069 |
const IMAGE * img2, |
const IMAGE * img2, |
1102 |
if(w > width) |
if(w > width) |
1103 |
w = width; |
w = width; |
1104 |
w2 = w >> 1; |
w2 = w >> 1; |
1105 |
void __inline |
|
1106 |
dY = (uint8_t*)out_frm->y + (mby << 4) * out_frm->stride_y + (mbx << 4); |
dY = (uint8_t*)out_frm->y + (mby << 4) * out_frm->stride_y + (mbx << 4); |
1107 |
dU = (uint8_t*)out_frm->u + (mby << 3) * out_frm->stride_u + (mbx << 3); |
dU = (uint8_t*)out_frm->u + (mby << 3) * out_frm->stride_u + (mbx << 3); |
1108 |
dV = (uint8_t*)out_frm->v + (mby << 3) * out_frm->stride_v + (mbx << 3); |
dV = (uint8_t*)out_frm->v + (mby << 3) * out_frm->stride_v + (mbx << 3); |
1112 |
|
|
1113 |
for(i = 0 ; i < 16 ; i++) { |
for(i = 0 ; i < 16 ; i++) { |
1114 |
memcpy(dY,sY,w); |
memcpy(dY,sY,w); |
1115 |
dY = out_frm->y + (mby << 4) * out_frm->stride_y + (mbx << 4); |
dY += out_frm->stride_y; |
1116 |
dU = out_frm->u + (mby << 3) * out_frm->stride_u + (mbx << 3); |
sY += std; |
1117 |
dV = out_frm->v + (mby << 3) * out_frm->stride_v + (mbx << 3); |
} |
1118 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
1119 |
memcpy(dU,sU,w2); |
memcpy(dU,sU,w2); |
1120 |
dU += out_frm->stride_u; |
dU += out_frm->stride_u; |
1126 |
sV += std2; |
sV += std2; |
1127 |
} |
} |
1128 |
} |
} |
1129 |
|
|
1130 |
|
|
1131 |
|
void |
1132 |
|
image_clear(IMAGE * img, int width, int height, int edged_width, |
1133 |
|
int y, int u, int v) |
1134 |
|
{ |
1135 |
|
uint8_t * p; |
1136 |
|
int i; |
1137 |
|
|
1138 |
|
p = img->y; |
1139 |
|
for (i = 0; i < height; i++) { |
1140 |
|
memset(p, y, width); |
1141 |
|
p += edged_width; |
1142 |
|
} |
1143 |
|
|
1144 |
|
p = img->u; |
1145 |
|
for (i = 0; i < height/2; i++) { |
1146 |
|
memset(p, u, width/2); |
1147 |
|
p += edged_width/2; |
1148 |
|
} |
1149 |
|
|
1150 |
|
p = img->v; |
1151 |
|
for (i = 0; i < height/2; i++) { |
1152 |
|
memset(p, v, width/2); |
1153 |
|
p += edged_width/2; |
1154 |
|
} |
1155 |
|
} |
1156 |
|
|
1157 |
|
|
1158 |
|
/* reduced resolution deblocking filter |
1159 |
|
block = block size (16=rrv, 8=full resolution) |
1160 |
|
flags = XVID_DEC_YDEBLOCK|XVID_DEC_UVDEBLOCK |
1161 |
|
*/ |
1162 |
|
void |
1163 |
|
image_deblock_rrv(IMAGE * img, int edged_width, |
1164 |
|
const MACROBLOCK * mbs, int mb_width, int mb_height, int mb_stride, |
1165 |
|
int block, int flags) |
1166 |
|
{ |
1167 |
|
const int edged_width2 = edged_width /2; |
1168 |
|
const int nblocks = block / 8; /* skals code uses 8pixel block uints */ |
1169 |
|
int i,j; |
1170 |
|
|
1171 |
|
/* luma: j,i in block units */ |
1172 |
|
if ((flags & XVID_DEC_DEBLOCKY)) |
1173 |
|
{ |
1174 |
|
for (j = 1; j < mb_height*2; j++) /* horizontal deblocking */ |
1175 |
|
for (i = 0; i < mb_width*2; i++) |
1176 |
|
{ |
1177 |
|
if (mbs[(j-1)/2*mb_stride + (i/2)].mode != MODE_NOT_CODED || |
1178 |
|
mbs[(j+0)/2*mb_stride + (i/2)].mode != MODE_NOT_CODED) |
1179 |
|
{ |
1180 |
|
hfilter_31(img->y + (j*block - 1)*edged_width + i*block, |
1181 |
|
img->y + (j*block + 0)*edged_width + i*block, nblocks); |
1182 |
|
} |
1183 |
|
} |
1184 |
|
|
1185 |
|
for (j = 0; j < mb_height*2; j++) /* vertical deblocking */ |
1186 |
|
for (i = 1; i < mb_width*2; i++) |
1187 |
|
{ |
1188 |
|
if (mbs[(j/2)*mb_stride + (i-1)/2].mode != MODE_NOT_CODED || |
1189 |
|
mbs[(j/2)*mb_stride + (i+0)/2].mode != MODE_NOT_CODED) |
1190 |
|
{ |
1191 |
|
vfilter_31(img->y + (j*block)*edged_width + i*block - 1, |
1192 |
|
img->y + (j*block)*edged_width + i*block + 0, |
1193 |
|
edged_width, nblocks); |
1194 |
|
} |
1195 |
|
} |
1196 |
|
} |
1197 |
|
|
1198 |
|
|
1199 |
|
/* chroma */ |
1200 |
|
if ((flags & XVID_DEC_DEBLOCKUV)) |
1201 |
|
{ |
1202 |
|
for (j = 1; j < mb_height; j++) /* horizontal deblocking */ |
1203 |
|
for (i = 0; i < mb_width; i++) |
1204 |
|
{ |
1205 |
|
if (mbs[(j-1)*mb_stride + i].mode != MODE_NOT_CODED || |
1206 |
|
mbs[(j+0)*mb_stride + i].mode != MODE_NOT_CODED) |
1207 |
|
{ |
1208 |
|
hfilter_31(img->u + (j*block - 1)*edged_width2 + i*block, |
1209 |
|
img->u + (j*block + 0)*edged_width2 + i*block, nblocks); |
1210 |
|
hfilter_31(img->v + (j*block - 1)*edged_width2 + i*block, |
1211 |
|
img->v + (j*block + 0)*edged_width2 + i*block, nblocks); |
1212 |
|
} |
1213 |
|
} |
1214 |
|
|
1215 |
|
for (j = 0; j < mb_height; j++) /* vertical deblocking */ |
1216 |
|
for (i = 1; i < mb_width; i++) |
1217 |
|
{ |
1218 |
|
if (mbs[j*mb_stride + i - 1].mode != MODE_NOT_CODED || |
1219 |
|
mbs[j*mb_stride + i + 0].mode != MODE_NOT_CODED) |
1220 |
|
{ |
1221 |
|
vfilter_31(img->u + (j*block)*edged_width2 + i*block - 1, |
1222 |
|
img->u + (j*block)*edged_width2 + i*block + 0, |
1223 |
|
edged_width2, nblocks); |
1224 |
|
vfilter_31(img->v + (j*block)*edged_width2 + i*block - 1, |
1225 |
|
img->v + (j*block)*edged_width2 + i*block + 0, |
1226 |
|
edged_width2, nblocks); |
1227 |
|
} |
1228 |
|
} |
1229 |
|
} |
1230 |
|
|
1231 |
|
} |
1232 |
|
|