3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Image management functions - |
* - Image management functions - |
5 |
* |
* |
6 |
* Copyright(C) 2001-2003 Peter Ross <pross@xvid.org> |
* Copyright(C) 2001-2004 Peter Ross <pross@xvid.org> |
7 |
* |
* |
8 |
* This program is free software ; you can redistribute it and/or modify |
* This program is free software ; you can redistribute it and/or modify |
9 |
* 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 |
26 |
#include <stdlib.h> |
#include <stdlib.h> |
27 |
#include <string.h> /* memcpy, memset */ |
#include <string.h> /* memcpy, memset */ |
28 |
#include <math.h> |
#include <math.h> |
|
|
|
29 |
#include "../portab.h" |
#include "../portab.h" |
30 |
#include "../global.h" /* XVID_CSP_XXX's */ |
#include "../global.h" /* XVID_CSP_XXX's */ |
31 |
#include "../xvid.h" /* XVID_CSP_XXX's */ |
#include "../xvid.h" /* XVID_CSP_XXX's */ |
32 |
#include "image.h" |
#include "image.h" |
33 |
#include "colorspace.h" |
#include "colorspace.h" |
34 |
#include "interpolate8x8.h" |
#include "interpolate8x8.h" |
|
#include "reduced.h" |
|
35 |
#include "../utils/mem_align.h" |
#include "../utils/mem_align.h" |
36 |
|
#include "../motion/sad.h" |
37 |
|
#include "../utils/emms.h" |
38 |
|
|
39 |
#include "font.h" /* XXX: remove later */ |
#include "font.h" /* XXX: remove later */ |
40 |
|
|
237 |
} |
} |
238 |
} |
} |
239 |
|
|
|
/* bframe encoding requires image-based u,v interpolation */ |
|
240 |
void |
void |
241 |
image_interpolate(const IMAGE * refn, |
image_interpolate(const uint8_t * refn, |
242 |
IMAGE * refh, |
uint8_t * refh, |
243 |
IMAGE * refv, |
uint8_t * refv, |
244 |
IMAGE * refhv, |
uint8_t * refhv, |
245 |
uint32_t edged_width, |
uint32_t edged_width, |
246 |
uint32_t edged_height, |
uint32_t edged_height, |
247 |
uint32_t quarterpel, |
uint32_t quarterpel, |
249 |
{ |
{ |
250 |
const uint32_t offset = EDGE_SIZE2 * (edged_width + 1); /* we only interpolate half of the edge area */ |
const uint32_t offset = EDGE_SIZE2 * (edged_width + 1); /* we only interpolate half of the edge area */ |
251 |
const uint32_t stride_add = 7 * edged_width; |
const uint32_t stride_add = 7 * edged_width; |
|
#if 0 |
|
|
const uint32_t edged_width2 = edged_width / 2; |
|
|
const uint32_t edged_height2 = edged_height / 2; |
|
|
const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); |
|
|
const uint32_t stride_add2 = 7 * edged_width2; |
|
|
#endif |
|
|
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
|
|
uint32_t x, y; |
|
252 |
|
|
253 |
|
uint8_t *n_ptr; |
254 |
|
uint8_t *h_ptr, *v_ptr, *hv_ptr; |
255 |
|
uint32_t x, y; |
256 |
|
|
257 |
n_ptr = refn->y; |
n_ptr = (uint8_t*)refn; |
258 |
h_ptr = refh->y; |
h_ptr = refh; |
259 |
v_ptr = refv->y; |
v_ptr = refv; |
260 |
|
|
261 |
n_ptr -= offset; |
n_ptr -= offset; |
262 |
h_ptr -= offset; |
h_ptr -= offset; |
285 |
n_ptr += stride_add; |
n_ptr += stride_add; |
286 |
} |
} |
287 |
|
|
288 |
h_ptr = refh->y + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; |
h_ptr = refh + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; |
289 |
hv_ptr = refhv->y + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; |
hv_ptr = refhv + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; |
290 |
|
|
291 |
for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { |
292 |
hv_ptr -= stride_add; |
hv_ptr -= stride_add; |
302 |
} |
} |
303 |
} else { |
} else { |
304 |
|
|
305 |
hv_ptr = refhv->y; |
hv_ptr = refhv; |
306 |
hv_ptr -= offset; |
hv_ptr -= offset; |
307 |
|
|
308 |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
328 |
n_ptr += stride_add; |
n_ptr += stride_add; |
329 |
} |
} |
330 |
} |
} |
|
/* |
|
|
#ifdef BFRAMES |
|
|
n_ptr = refn->u; |
|
|
h_ptr = refh->u; |
|
|
v_ptr = refv->u; |
|
|
hv_ptr = refhv->u; |
|
|
|
|
|
n_ptr -= offset2; |
|
|
h_ptr -= offset2; |
|
|
v_ptr -= offset2; |
|
|
hv_ptr -= offset2; |
|
|
|
|
|
for (y = 0; y < edged_height2; y += 8) { |
|
|
for (x = 0; x < edged_width2; x += 8) { |
|
|
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width2, rounding); |
|
|
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width2, rounding); |
|
|
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width2, rounding); |
|
|
|
|
|
n_ptr += 8; |
|
|
h_ptr += 8; |
|
|
v_ptr += 8; |
|
|
hv_ptr += 8; |
|
|
} |
|
|
h_ptr += stride_add2; |
|
|
v_ptr += stride_add2; |
|
|
hv_ptr += stride_add2; |
|
|
n_ptr += stride_add2; |
|
|
} |
|
|
|
|
|
n_ptr = refn->v; |
|
|
h_ptr = refh->v; |
|
|
v_ptr = refv->v; |
|
|
hv_ptr = refhv->v; |
|
|
|
|
|
n_ptr -= offset2; |
|
|
h_ptr -= offset2; |
|
|
v_ptr -= offset2; |
|
|
hv_ptr -= offset2; |
|
|
|
|
|
for (y = 0; y < edged_height2; y = y + 8) { |
|
|
for (x = 0; x < edged_width2; x = x + 8) { |
|
|
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width2, rounding); |
|
|
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width2, rounding); |
|
|
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width2, rounding); |
|
|
|
|
|
n_ptr += 8; |
|
|
h_ptr += 8; |
|
|
v_ptr += 8; |
|
|
hv_ptr += 8; |
|
|
} |
|
|
h_ptr += stride_add2; |
|
|
v_ptr += stride_add2; |
|
|
hv_ptr += stride_add2; |
|
|
n_ptr += stride_add2; |
|
|
} |
|
|
#endif |
|
|
*/ |
|
|
/* |
|
|
interpolate_halfpel_h( |
|
|
refh->y - offset, |
|
|
refn->y - offset, |
|
|
edged_width, edged_height, |
|
|
rounding); |
|
|
|
|
|
interpolate_halfpel_v( |
|
|
refv->y - offset, |
|
|
refn->y - offset, |
|
|
edged_width, edged_height, |
|
|
rounding); |
|
|
|
|
|
interpolate_halfpel_hv( |
|
|
refhv->y - offset, |
|
|
refn->y - offset, |
|
|
edged_width, edged_height, |
|
|
rounding); |
|
|
*/ |
|
|
|
|
|
/* uv-image-based compensation |
|
|
offset = EDGE_SIZE2 * (edged_width / 2 + 1); |
|
|
|
|
|
interpolate_halfpel_h( |
|
|
refh->u - offset, |
|
|
refn->u - offset, |
|
|
edged_width / 2, edged_height / 2, |
|
|
rounding); |
|
|
|
|
|
interpolate_halfpel_v( |
|
|
refv->u - offset, |
|
|
refn->u - offset, |
|
|
edged_width / 2, edged_height / 2, |
|
|
rounding); |
|
|
|
|
|
interpolate_halfpel_hv( |
|
|
refhv->u - offset, |
|
|
refn->u - offset, |
|
|
edged_width / 2, edged_height / 2, |
|
|
rounding); |
|
|
|
|
|
|
|
|
interpolate_halfpel_h( |
|
|
refh->v - offset, |
|
|
refn->v - offset, |
|
|
edged_width / 2, edged_height / 2, |
|
|
rounding); |
|
|
|
|
|
interpolate_halfpel_v( |
|
|
refv->v - offset, |
|
|
refn->v - offset, |
|
|
edged_width / 2, edged_height / 2, |
|
|
rounding); |
|
|
|
|
|
interpolate_halfpel_hv( |
|
|
refhv->v - offset, |
|
|
refn->v - offset, |
|
|
edged_width / 2, edged_height / 2, |
|
|
rounding); |
|
|
*/ |
|
331 |
} |
} |
332 |
|
|
333 |
|
|
385 |
uint8_t * y_ptr, uint8_t * u_ptr, uint8_t * v_ptr, |
uint8_t * y_ptr, uint8_t * u_ptr, uint8_t * v_ptr, |
386 |
int y_stride, int uv_stride, |
int y_stride, int uv_stride, |
387 |
int width, int height, int vflip, |
int width, int height, int vflip, |
388 |
packedFunc * func_opt, packedFunc func_c, int size) |
packedFunc * func_opt, packedFunc func_c, |
389 |
|
int size, int interlacing) |
390 |
{ |
{ |
391 |
int width_opt, width_c; |
int width_opt, width_c, height_opt; |
392 |
|
|
393 |
|
if (width<0 || width==1 || height==1) return; /* forget about it */ |
394 |
|
|
395 |
if (func_opt != func_c && x_stride < size*((width+15)/16)*16) |
if (func_opt != func_c && x_stride < size*((width+15)/16)*16) |
396 |
{ |
{ |
397 |
width_opt = width & (~15); |
width_opt = width & (~15); |
398 |
width_c = width - width_opt; |
width_c = (width - width_opt) & (~1); |
399 |
} |
} |
400 |
else |
else if (func_opt != func_c && !(width&1) && (size==3)) |
401 |
{ |
{ |
402 |
width_opt = width; |
/* MMX reads 4 bytes per pixel for RGB/BGR */ |
403 |
|
width_opt = width - 2; |
404 |
|
width_c = 2; |
405 |
|
} |
406 |
|
else { |
407 |
|
/* Enforce the width to be divisable by two. */ |
408 |
|
width_opt = width & (~1); |
409 |
width_c = 0; |
width_c = 0; |
410 |
} |
} |
411 |
|
|
412 |
|
/* packed conversions require height to be divisable by 2 |
413 |
|
(or even by 4 for interlaced conversion) */ |
414 |
|
if (interlacing) |
415 |
|
height_opt = height & (~3); |
416 |
|
else |
417 |
|
height_opt = height & (~1); |
418 |
|
|
419 |
func_opt(x_ptr, x_stride, |
func_opt(x_ptr, x_stride, |
420 |
y_ptr, u_ptr, v_ptr, y_stride, uv_stride, |
y_ptr, u_ptr, v_ptr, y_stride, uv_stride, |
421 |
width_opt, height, vflip); |
width_opt, height_opt, vflip); |
422 |
|
|
423 |
if (width_c) |
if (width_c) |
424 |
{ |
{ |
425 |
func_c(x_ptr + size*width_opt, x_stride, |
func_c(x_ptr + size*width_opt, x_stride, |
426 |
y_ptr + width_opt, u_ptr + width_opt/2, v_ptr + width_opt/2, |
y_ptr + width_opt, u_ptr + width_opt/2, v_ptr + width_opt/2, |
427 |
y_stride, uv_stride, width_c, height, vflip); |
y_stride, uv_stride, width_c, height_opt, vflip); |
428 |
} |
} |
429 |
} |
} |
430 |
|
|
453 |
src[0], src_stride[0], image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
454 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
455 |
interlacing?rgb555i_to_yv12 :rgb555_to_yv12, |
interlacing?rgb555i_to_yv12 :rgb555_to_yv12, |
456 |
interlacing?rgb555i_to_yv12_c:rgb555_to_yv12_c, 2); |
interlacing?rgb555i_to_yv12_c:rgb555_to_yv12_c, 2, interlacing); |
457 |
break; |
break; |
458 |
|
|
459 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
461 |
src[0], src_stride[0], image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
462 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
463 |
interlacing?rgb565i_to_yv12 :rgb565_to_yv12, |
interlacing?rgb565i_to_yv12 :rgb565_to_yv12, |
464 |
interlacing?rgb565i_to_yv12_c:rgb565_to_yv12_c, 2); |
interlacing?rgb565i_to_yv12_c:rgb565_to_yv12_c, 2, interlacing); |
465 |
break; |
break; |
466 |
|
|
467 |
|
|
470 |
src[0], src_stride[0], image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
471 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
472 |
interlacing?bgri_to_yv12 :bgr_to_yv12, |
interlacing?bgri_to_yv12 :bgr_to_yv12, |
473 |
interlacing?bgri_to_yv12_c:bgr_to_yv12_c, 3); |
interlacing?bgri_to_yv12_c:bgr_to_yv12_c, 3, interlacing); |
474 |
break; |
break; |
475 |
|
|
476 |
case XVID_CSP_BGRA: |
case XVID_CSP_BGRA: |
478 |
src[0], src_stride[0], image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
479 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
480 |
interlacing?bgrai_to_yv12 :bgra_to_yv12, |
interlacing?bgrai_to_yv12 :bgra_to_yv12, |
481 |
interlacing?bgrai_to_yv12_c:bgra_to_yv12_c, 4); |
interlacing?bgrai_to_yv12_c:bgra_to_yv12_c, 4, interlacing); |
482 |
break; |
break; |
483 |
|
|
484 |
case XVID_CSP_ABGR : |
case XVID_CSP_ABGR : |
486 |
src[0], src_stride[0], image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
487 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
488 |
interlacing?abgri_to_yv12 :abgr_to_yv12, |
interlacing?abgri_to_yv12 :abgr_to_yv12, |
489 |
interlacing?abgri_to_yv12_c:abgr_to_yv12_c, 4); |
interlacing?abgri_to_yv12_c:abgr_to_yv12_c, 4, interlacing); |
490 |
|
break; |
491 |
|
|
492 |
|
case XVID_CSP_RGB: |
493 |
|
safe_packed_conv( |
494 |
|
src[0], src_stride[0], image->y, image->u, image->v, |
495 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
496 |
|
interlacing?rgbi_to_yv12 :rgb_to_yv12, |
497 |
|
interlacing?rgbi_to_yv12_c:rgb_to_yv12_c, 3, interlacing); |
498 |
break; |
break; |
499 |
|
|
500 |
case XVID_CSP_RGBA : |
case XVID_CSP_RGBA : |
502 |
src[0], src_stride[0], image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
503 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
504 |
interlacing?rgbai_to_yv12 :rgba_to_yv12, |
interlacing?rgbai_to_yv12 :rgba_to_yv12, |
505 |
interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4); |
interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4, interlacing); |
506 |
break; |
break; |
507 |
|
|
508 |
case XVID_CSP_ARGB: |
case XVID_CSP_ARGB: |
510 |
src[0], src_stride[0], image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
511 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
512 |
interlacing?argbi_to_yv12 : argb_to_yv12, |
interlacing?argbi_to_yv12 : argb_to_yv12, |
513 |
interlacing?argbi_to_yv12_c: argb_to_yv12_c, 4); |
interlacing?argbi_to_yv12_c: argb_to_yv12_c, 4, interlacing); |
514 |
break; |
break; |
515 |
|
|
516 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
518 |
src[0], src_stride[0], image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
519 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
520 |
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
521 |
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2, interlacing); |
522 |
break; |
break; |
523 |
|
|
524 |
case XVID_CSP_YVYU: /* u/v swapped */ |
case XVID_CSP_YVYU: /* u/v swapped */ |
526 |
src[0], src_stride[0], image->y, image->v, image->u, |
src[0], src_stride[0], image->y, image->v, image->u, |
527 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
528 |
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
529 |
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2, interlacing); |
530 |
break; |
break; |
531 |
|
|
532 |
case XVID_CSP_UYVY: |
case XVID_CSP_UYVY: |
534 |
src[0], src_stride[0], image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
535 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
536 |
interlacing?uyvyi_to_yv12 :uyvy_to_yv12, |
interlacing?uyvyi_to_yv12 :uyvy_to_yv12, |
537 |
interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2); |
interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2, interlacing); |
538 |
break; |
break; |
539 |
|
|
540 |
case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
619 |
int height, |
int height, |
620 |
uint32_t edged_width, |
uint32_t edged_width, |
621 |
uint8_t * dst[4], |
uint8_t * dst[4], |
622 |
uint32_t dst_stride[4], |
int dst_stride[4], |
623 |
int csp, |
int csp, |
624 |
int interlacing) |
int interlacing) |
625 |
{ |
{ |
638 |
dst[0], dst_stride[0], image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
639 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
640 |
interlacing?yv12_to_rgb555i :yv12_to_rgb555, |
interlacing?yv12_to_rgb555i :yv12_to_rgb555, |
641 |
interlacing?yv12_to_rgb555i_c:yv12_to_rgb555_c, 2); |
interlacing?yv12_to_rgb555i_c:yv12_to_rgb555_c, 2, interlacing); |
642 |
return 0; |
return 0; |
643 |
|
|
644 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
646 |
dst[0], dst_stride[0], image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
647 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
648 |
interlacing?yv12_to_rgb565i :yv12_to_rgb565, |
interlacing?yv12_to_rgb565i :yv12_to_rgb565, |
649 |
interlacing?yv12_to_rgb565i_c:yv12_to_rgb565_c, 2); |
interlacing?yv12_to_rgb565i_c:yv12_to_rgb565_c, 2, interlacing); |
650 |
return 0; |
return 0; |
651 |
|
|
652 |
case XVID_CSP_BGR: |
case XVID_CSP_BGR: |
654 |
dst[0], dst_stride[0], image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
655 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
656 |
interlacing?yv12_to_bgri :yv12_to_bgr, |
interlacing?yv12_to_bgri :yv12_to_bgr, |
657 |
interlacing?yv12_to_bgri_c:yv12_to_bgr_c, 3); |
interlacing?yv12_to_bgri_c:yv12_to_bgr_c, 3, interlacing); |
658 |
return 0; |
return 0; |
659 |
|
|
660 |
case XVID_CSP_BGRA: |
case XVID_CSP_BGRA: |
662 |
dst[0], dst_stride[0], image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
663 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
664 |
interlacing?yv12_to_bgrai :yv12_to_bgra, |
interlacing?yv12_to_bgrai :yv12_to_bgra, |
665 |
interlacing?yv12_to_bgrai_c:yv12_to_bgra_c, 4); |
interlacing?yv12_to_bgrai_c:yv12_to_bgra_c, 4, interlacing); |
666 |
return 0; |
return 0; |
667 |
|
|
668 |
case XVID_CSP_ABGR: |
case XVID_CSP_ABGR: |
670 |
dst[0], dst_stride[0], image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
671 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
672 |
interlacing?yv12_to_abgri :yv12_to_abgr, |
interlacing?yv12_to_abgri :yv12_to_abgr, |
673 |
interlacing?yv12_to_abgri_c:yv12_to_abgr_c, 4); |
interlacing?yv12_to_abgri_c:yv12_to_abgr_c, 4, interlacing); |
674 |
|
return 0; |
675 |
|
|
676 |
|
case XVID_CSP_RGB: |
677 |
|
safe_packed_conv( |
678 |
|
dst[0], dst_stride[0], image->y, image->u, image->v, |
679 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
680 |
|
interlacing?yv12_to_rgbi :yv12_to_rgb, |
681 |
|
interlacing?yv12_to_rgbi_c:yv12_to_rgb_c, 3, interlacing); |
682 |
return 0; |
return 0; |
683 |
|
|
684 |
case XVID_CSP_RGBA: |
case XVID_CSP_RGBA: |
686 |
dst[0], dst_stride[0], image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
687 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
688 |
interlacing?yv12_to_rgbai :yv12_to_rgba, |
interlacing?yv12_to_rgbai :yv12_to_rgba, |
689 |
interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4); |
interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4, interlacing); |
690 |
return 0; |
return 0; |
691 |
|
|
692 |
case XVID_CSP_ARGB: |
case XVID_CSP_ARGB: |
694 |
dst[0], dst_stride[0], image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
695 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
696 |
interlacing?yv12_to_argbi :yv12_to_argb, |
interlacing?yv12_to_argbi :yv12_to_argb, |
697 |
interlacing?yv12_to_argbi_c:yv12_to_argb_c, 4); |
interlacing?yv12_to_argbi_c:yv12_to_argb_c, 4, interlacing); |
698 |
return 0; |
return 0; |
699 |
|
|
700 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
702 |
dst[0], dst_stride[0], image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
703 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
704 |
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
705 |
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2, interlacing); |
706 |
return 0; |
return 0; |
707 |
|
|
708 |
case XVID_CSP_YVYU: /* u,v swapped */ |
case XVID_CSP_YVYU: /* u,v swapped */ |
710 |
dst[0], dst_stride[0], image->y, image->v, image->u, |
dst[0], dst_stride[0], image->y, image->v, image->u, |
711 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
712 |
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
713 |
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2, interlacing); |
714 |
return 0; |
return 0; |
715 |
|
|
716 |
case XVID_CSP_UYVY: |
case XVID_CSP_UYVY: |
718 |
dst[0], dst_stride[0], image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
719 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
720 |
interlacing?yv12_to_uyvyi :yv12_to_uyvy, |
interlacing?yv12_to_uyvyi :yv12_to_uyvy, |
721 |
interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2); |
interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2, interlacing); |
722 |
return 0; |
return 0; |
723 |
|
|
724 |
case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
808 |
uint16_t width, |
uint16_t width, |
809 |
uint16_t height) |
uint16_t height) |
810 |
{ |
{ |
811 |
int diff, x, y; |
int y, bwidth, bheight; |
812 |
long sse=0; |
long sse=0; |
813 |
|
|
814 |
for (y = 0; y < height; y++) { |
bwidth = width & (~0x07); |
815 |
|
bheight = height & (~0x07); |
816 |
|
|
817 |
|
/* Compute the 8x8 integer part */ |
818 |
|
for (y = 0; y<bheight; y += 8) { |
819 |
|
int x; |
820 |
|
|
821 |
|
/* Compute sse for the band */ |
822 |
|
for (x = 0; x<bwidth; x += 8) |
823 |
|
sse += sse8_8bit(orig + x, recon + x, stride); |
824 |
|
|
825 |
|
/* remaining pixels of the 8 pixels high band */ |
826 |
|
for (x = bwidth; x < width; x++) { |
827 |
|
int diff; |
828 |
|
diff = *(orig + 0*stride + x) - *(recon + 0*stride + x); |
829 |
|
sse += diff * diff; |
830 |
|
diff = *(orig + 1*stride + x) - *(recon + 1*stride + x); |
831 |
|
sse += diff * diff; |
832 |
|
diff = *(orig + 2*stride + x) - *(recon + 2*stride + x); |
833 |
|
sse += diff * diff; |
834 |
|
diff = *(orig + 3*stride + x) - *(recon + 3*stride + x); |
835 |
|
sse += diff * diff; |
836 |
|
diff = *(orig + 4*stride + x) - *(recon + 4*stride + x); |
837 |
|
sse += diff * diff; |
838 |
|
diff = *(orig + 5*stride + x) - *(recon + 5*stride + x); |
839 |
|
sse += diff * diff; |
840 |
|
diff = *(orig + 6*stride + x) - *(recon + 6*stride + x); |
841 |
|
sse += diff * diff; |
842 |
|
diff = *(orig + 7*stride + x) - *(recon + 7*stride + x); |
843 |
|
sse += diff * diff; |
844 |
|
} |
845 |
|
|
846 |
|
orig += 8*stride; |
847 |
|
recon += 8*stride; |
848 |
|
} |
849 |
|
|
850 |
|
/* Compute the down rectangle sse */ |
851 |
|
for (y = bheight; y < height; y++) { |
852 |
|
int x; |
853 |
for (x = 0; x < width; x++) { |
for (x = 0; x < width; x++) { |
854 |
|
int diff; |
855 |
diff = *(orig + x) - *(recon + x); |
diff = *(orig + x) - *(recon + x); |
856 |
sse += diff * diff; |
sse += diff * diff; |
857 |
} |
} |
858 |
orig += stride; |
orig += stride; |
859 |
recon += stride; |
recon += stride; |
860 |
} |
} |
861 |
return sse; |
|
862 |
|
return (sse); |
863 |
} |
} |
864 |
|
|
865 |
#if 0 |
#if 0 |
988 |
} |
} |
989 |
|
|
990 |
void |
void |
991 |
output_slice(IMAGE * cur, int std, int width, xvid_image_t* out_frm, int mbx, int mby,int mbl) { |
output_slice(IMAGE * cur, int stride, int width, xvid_image_t* out_frm, int mbx, int mby,int mbl) { |
992 |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
993 |
int std2 = std >> 1; |
int stride2 = stride >> 1; |
994 |
int w = mbl << 4, w2,i; |
int w = mbl << 4, w2,i; |
995 |
|
|
996 |
if(w > width) |
if(w > width) |
1000 |
dY = (uint8_t*)out_frm->plane[0] + (mby << 4) * out_frm->stride[0] + (mbx << 4); |
dY = (uint8_t*)out_frm->plane[0] + (mby << 4) * out_frm->stride[0] + (mbx << 4); |
1001 |
dU = (uint8_t*)out_frm->plane[1] + (mby << 3) * out_frm->stride[1] + (mbx << 3); |
dU = (uint8_t*)out_frm->plane[1] + (mby << 3) * out_frm->stride[1] + (mbx << 3); |
1002 |
dV = (uint8_t*)out_frm->plane[2] + (mby << 3) * out_frm->stride[2] + (mbx << 3); |
dV = (uint8_t*)out_frm->plane[2] + (mby << 3) * out_frm->stride[2] + (mbx << 3); |
1003 |
sY = cur->y + (mby << 4) * std + (mbx << 4); |
sY = cur->y + (mby << 4) * stride + (mbx << 4); |
1004 |
sU = cur->u + (mby << 3) * std2 + (mbx << 3); |
sU = cur->u + (mby << 3) * stride2 + (mbx << 3); |
1005 |
sV = cur->v + (mby << 3) * std2 + (mbx << 3); |
sV = cur->v + (mby << 3) * stride2 + (mbx << 3); |
1006 |
|
|
1007 |
for(i = 0 ; i < 16 ; i++) { |
for(i = 0 ; i < 16 ; i++) { |
1008 |
memcpy(dY,sY,w); |
memcpy(dY,sY,w); |
1009 |
dY += out_frm->stride[0]; |
dY += out_frm->stride[0]; |
1010 |
sY += std; |
sY += stride; |
1011 |
} |
} |
1012 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
1013 |
memcpy(dU,sU,w2); |
memcpy(dU,sU,w2); |
1014 |
dU += out_frm->stride[1]; |
dU += out_frm->stride[1]; |
1015 |
sU += std2; |
sU += stride2; |
1016 |
} |
} |
1017 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
1018 |
memcpy(dV,sV,w2); |
memcpy(dV,sV,w2); |
1019 |
dV += out_frm->stride[2]; |
dV += out_frm->stride[2]; |
1020 |
sV += std2; |
sV += stride2; |
1021 |
} |
} |
1022 |
} |
} |
1023 |
|
|
1048 |
} |
} |
1049 |
} |
} |
1050 |
|
|
1051 |
|
/****************************************************************************/ |
1052 |
|
|
1053 |
/* reduced resolution deblocking filter |
static void (*deintl_core)(uint8_t *, int width, int height, const int stride) = 0; |
1054 |
block = block size (16=rrv, 8=full resolution) |
extern void xvid_deinterlace_sse(uint8_t *, int width, int height, const int stride); |
|
flags = XVID_DEC_YDEBLOCK|XVID_DEC_UVDEBLOCK |
|
|
*/ |
|
|
void |
|
|
image_deblock_rrv(IMAGE * img, int edged_width, |
|
|
const MACROBLOCK * mbs, int mb_width, int mb_height, int mb_stride, |
|
|
int block, int flags) |
|
|
{ |
|
|
const int edged_width2 = edged_width /2; |
|
|
const int nblocks = block / 8; /* skals code uses 8pixel block uints */ |
|
|
int i,j; |
|
1055 |
|
|
1056 |
/* luma: j,i in block units */ |
#define CLIP_255(x) ( ((x)&~255) ? ((-(x)) >> (8*sizeof((x))-1))&0xff : (x) ) |
1057 |
|
|
1058 |
for (j = 1; j < mb_height*2; j++) /* horizontal deblocking */ |
static void deinterlace_c(uint8_t *pix, int width, int height, const int bps) |
|
for (i = 0; i < mb_width*2; i++) |
|
1059 |
{ |
{ |
1060 |
if (mbs[(j-1)/2*mb_stride + (i/2)].mode != MODE_NOT_CODED || |
pix += bps; |
1061 |
mbs[(j+0)/2*mb_stride + (i/2)].mode != MODE_NOT_CODED) |
while(width-->0) |
|
{ |
|
|
hfilter_31(img->y + (j*block - 1)*edged_width + i*block, |
|
|
img->y + (j*block + 0)*edged_width + i*block, nblocks); |
|
|
} |
|
|
} |
|
|
|
|
|
for (j = 0; j < mb_height*2; j++) /* vertical deblocking */ |
|
|
for (i = 1; i < mb_width*2; i++) |
|
1062 |
{ |
{ |
1063 |
if (mbs[(j/2)*mb_stride + (i-1)/2].mode != MODE_NOT_CODED || |
int p1 = pix[-bps]; |
1064 |
mbs[(j/2)*mb_stride + (i+0)/2].mode != MODE_NOT_CODED) |
int p2 = pix[0]; |
1065 |
|
int p0 = p2; |
1066 |
|
int j = (height>>1) - 1; |
1067 |
|
int V; |
1068 |
|
unsigned char *P = pix++; |
1069 |
|
while(j-->0) |
1070 |
{ |
{ |
1071 |
vfilter_31(img->y + (j*block)*edged_width + i*block - 1, |
const int p3 = P[ bps]; |
1072 |
img->y + (j*block)*edged_width + i*block + 0, |
const int p4 = P[2*bps]; |
1073 |
edged_width, nblocks); |
V = ((p1+p3+1)>>1) + ((p2 - ((p0+p4+1)>>1)) >> 2); |
1074 |
|
P[0] = CLIP_255( V ); |
1075 |
|
p0 = p2; |
1076 |
|
p1 = p3; |
1077 |
|
p2 = p4; |
1078 |
|
P += 2*bps; |
1079 |
} |
} |
1080 |
} |
V = ((p1+p1+1)>>1) + ((p2 - ((p0+p2+1)>>1)) >> 2); |
1081 |
|
P[0] = CLIP_255( V ); |
|
|
|
|
|
|
|
/* chroma */ |
|
|
|
|
|
for (j = 1; j < mb_height; j++) /* horizontal deblocking */ |
|
|
for (i = 0; i < mb_width; i++) |
|
|
{ |
|
|
if (mbs[(j-1)*mb_stride + i].mode != MODE_NOT_CODED || |
|
|
mbs[(j+0)*mb_stride + i].mode != MODE_NOT_CODED) |
|
|
{ |
|
|
hfilter_31(img->u + (j*block - 1)*edged_width2 + i*block, |
|
|
img->u + (j*block + 0)*edged_width2 + i*block, nblocks); |
|
|
hfilter_31(img->v + (j*block - 1)*edged_width2 + i*block, |
|
|
img->v + (j*block + 0)*edged_width2 + i*block, nblocks); |
|
1082 |
} |
} |
1083 |
} |
} |
1084 |
|
#undef CLIP_255 |
1085 |
|
|
1086 |
for (j = 0; j < mb_height; j++) /* vertical deblocking */ |
int xvid_image_deinterlace(xvid_image_t* img, int width, int height, int bottom_first) |
|
for (i = 1; i < mb_width; i++) |
|
|
{ |
|
|
if (mbs[j*mb_stride + i - 1].mode != MODE_NOT_CODED || |
|
|
mbs[j*mb_stride + i + 0].mode != MODE_NOT_CODED) |
|
1087 |
{ |
{ |
1088 |
vfilter_31(img->u + (j*block)*edged_width2 + i*block - 1, |
if (height&1) |
1089 |
img->u + (j*block)*edged_width2 + i*block + 0, |
return 0; |
1090 |
edged_width2, nblocks); |
if (img->csp!=XVID_CSP_PLANAR && img->csp!=XVID_CSP_I420 && img->csp!=XVID_CSP_YV12) |
1091 |
vfilter_31(img->v + (j*block)*edged_width2 + i*block - 1, |
return 0; /* not yet supported */ |
1092 |
img->v + (j*block)*edged_width2 + i*block + 0, |
if (deintl_core==0) { |
1093 |
edged_width2, nblocks); |
deintl_core = deinterlace_c; |
1094 |
|
#if defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) |
1095 |
|
{ |
1096 |
|
int cpu_flags = check_cpu_features(); |
1097 |
|
if (cpu_flags & XVID_CPU_MMX) |
1098 |
|
deintl_core = xvid_deinterlace_sse; |
1099 |
|
} |
1100 |
|
#endif |
1101 |
} |
} |
1102 |
|
if (!bottom_first) { |
1103 |
|
deintl_core(img->plane[0], width, height, img->stride[0]); |
1104 |
|
deintl_core(img->plane[1], width>>1, height>>1, img->stride[1]); |
1105 |
|
deintl_core(img->plane[2], width>>1, height>>1, img->stride[2]); |
1106 |
|
} |
1107 |
|
else { |
1108 |
|
deintl_core((uint8_t *)img->plane[0] + ( height -1)*img->stride[0], width, height, -img->stride[0]); |
1109 |
|
deintl_core((uint8_t *)img->plane[1] + ((height>>1)-1)*img->stride[1], width>>1, height>>1, -img->stride[1]); |
1110 |
|
deintl_core((uint8_t *)img->plane[2] + ((height>>1)-1)*img->stride[2], width>>1, height>>1, -img->stride[2]); |
1111 |
} |
} |
1112 |
|
emms(); |
1113 |
|
|
1114 |
|
return 1; |
1115 |
} |
} |
1116 |
|
|