1 |
/***************************************************************************** |
/************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - image module - |
* - Image management functions - |
5 |
* |
* |
6 |
* Copyright(C) 2002 Peter Ross <pross@xvid.org> |
* Copyright(C) 2001-2010 Peter Ross <pross@xvid.org> |
7 |
* |
* |
8 |
* This file is part of XviD, a free MPEG-4 video encoder/decoder |
* 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 |
|
* XviD is free software; you can redistribute it and/or modify it |
|
|
* under the terms of the GNU General Public License as published by |
|
10 |
* the Free Software Foundation; either version 2 of the License, or |
* the Free Software Foundation; either version 2 of the License, or |
11 |
* (at your option) any later version. |
* (at your option) any later version. |
12 |
* |
* |
19 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
20 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
21 |
* |
* |
|
* Under section 8 of the GNU General Public License, the copyright |
|
|
* holders of XVID explicitly forbid distribution in the following |
|
|
* countries: |
|
|
* |
|
|
* - Japan |
|
|
* - United States of America |
|
|
* |
|
|
* Linking XviD statically or dynamically with other modules is making a |
|
|
* combined work based on XviD. Thus, the terms and conditions of the |
|
|
* GNU General Public License cover the whole combination. |
|
|
* |
|
|
* As a special exception, the copyright holders of XviD give you |
|
|
* permission to link XviD with independent modules that communicate with |
|
|
* XviD solely through the VFW1.1 and DShow interfaces, regardless of the |
|
|
* license terms of these independent modules, and to copy and distribute |
|
|
* the resulting combined work under terms of your choice, provided that |
|
|
* every copy of the combined work is accompanied by a complete copy of |
|
|
* the source code of XviD (the version of XviD used to produce the |
|
|
* combined work), being distributed under the terms of the GNU General |
|
|
* Public License plus this exception. An independent module is a module |
|
|
* which is not derived from or based on XviD. |
|
|
* |
|
|
* Note that people who make modified versions of XviD are not obligated |
|
|
* to grant this special exception for their modified versions; it is |
|
|
* their choice whether to do so. The GNU General Public License gives |
|
|
* permission to release a modified version without this exception; this |
|
|
* exception also makes it possible to release a modified version which |
|
|
* carries forward this exception. |
|
|
* |
|
22 |
* $Id$ |
* $Id$ |
23 |
* |
* |
24 |
****************************************************************************/ |
****************************************************************************/ |
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 */ |
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 "../divx4.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 */ |
40 |
|
|
41 |
#define SAFETY 64 |
#define SAFETY 64 |
42 |
#define EDGE_SIZE2 (EDGE_SIZE/2) |
#define EDGE_SIZE2 (EDGE_SIZE/2) |
49 |
{ |
{ |
50 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
51 |
const uint32_t edged_height2 = edged_height / 2; |
const uint32_t edged_height2 = edged_height / 2; |
|
uint32_t i; |
|
52 |
|
|
53 |
image->y = |
image->y = |
54 |
xvid_malloc(edged_width * (edged_height + 1) + SAFETY, CACHE_LINE); |
xvid_malloc(edged_width * (edged_height + 1) + SAFETY, CACHE_LINE); |
55 |
if (image->y == NULL) { |
if (image->y == NULL) { |
56 |
return -1; |
return -1; |
57 |
} |
} |
58 |
|
memset(image->y, 0, edged_width * (edged_height + 1) + SAFETY); |
|
for (i = 0; i < edged_width * edged_height + SAFETY; i++) { |
|
|
image->y[i] = 0; |
|
|
} |
|
59 |
|
|
60 |
image->u = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
image->u = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
61 |
if (image->u == NULL) { |
if (image->u == NULL) { |
62 |
xvid_free(image->y); |
xvid_free(image->y); |
63 |
|
image->y = NULL; |
64 |
return -1; |
return -1; |
65 |
} |
} |
66 |
|
memset(image->u, 0, edged_width2 * edged_height2 + SAFETY); |
67 |
|
|
68 |
image->v = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
image->v = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
69 |
if (image->v == NULL) { |
if (image->v == NULL) { |
70 |
xvid_free(image->u); |
xvid_free(image->u); |
71 |
|
image->u = NULL; |
72 |
xvid_free(image->y); |
xvid_free(image->y); |
73 |
|
image->y = NULL; |
74 |
return -1; |
return -1; |
75 |
} |
} |
76 |
|
memset(image->v, 0, edged_width2 * edged_height2 + SAFETY); |
77 |
|
|
78 |
image->y += EDGE_SIZE * edged_width + EDGE_SIZE; |
image->y += EDGE_SIZE * edged_width + EDGE_SIZE; |
79 |
image->u += EDGE_SIZE2 * edged_width2 + EDGE_SIZE2; |
image->u += EDGE_SIZE2 * edged_width2 + EDGE_SIZE2; |
93 |
|
|
94 |
if (image->y) { |
if (image->y) { |
95 |
xvid_free(image->y - (EDGE_SIZE * edged_width + EDGE_SIZE)); |
xvid_free(image->y - (EDGE_SIZE * edged_width + EDGE_SIZE)); |
96 |
|
image->y = NULL; |
97 |
} |
} |
98 |
if (image->u) { |
if (image->u) { |
99 |
xvid_free(image->u - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
xvid_free(image->u - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
100 |
|
image->u = NULL; |
101 |
} |
} |
102 |
if (image->v) { |
if (image->v) { |
103 |
xvid_free(image->v - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
xvid_free(image->v - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
104 |
|
image->v = NULL; |
105 |
} |
} |
106 |
} |
} |
107 |
|
|
110 |
image_swap(IMAGE * image1, |
image_swap(IMAGE * image1, |
111 |
IMAGE * image2) |
IMAGE * image2) |
112 |
{ |
{ |
113 |
uint8_t *tmp; |
SWAP(uint8_t*, image1->y, image2->y); |
114 |
|
SWAP(uint8_t*, image1->u, image2->u); |
115 |
tmp = image1->y; |
SWAP(uint8_t*, image1->v, image2->v); |
|
image1->y = image2->y; |
|
|
image2->y = tmp; |
|
|
|
|
|
tmp = image1->u; |
|
|
image1->u = image2->u; |
|
|
image2->u = tmp; |
|
|
|
|
|
tmp = image1->v; |
|
|
image1->v = image2->v; |
|
|
image2->v = tmp; |
|
116 |
} |
} |
117 |
|
|
118 |
|
|
127 |
memcpy(image1->v, image2->v, edged_width * height / 4); |
memcpy(image1->v, image2->v, edged_width * height / 4); |
128 |
} |
} |
129 |
|
|
130 |
|
/* setedges bug was fixed in this BS version */ |
131 |
|
#define SETEDGES_BUG_BEFORE 18 |
132 |
|
|
133 |
void |
void |
134 |
image_setedges(IMAGE * image, |
image_setedges(IMAGE * image, |
135 |
uint32_t edged_width, |
uint32_t edged_width, |
136 |
uint32_t edged_height, |
uint32_t edged_height, |
137 |
uint32_t width, |
uint32_t width, |
138 |
uint32_t height) |
uint32_t height, |
139 |
|
int bs_version) |
140 |
{ |
{ |
141 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
142 |
const uint32_t width2 = width / 2; |
uint32_t width2; |
143 |
uint32_t i; |
uint32_t i; |
144 |
uint8_t *dst; |
uint8_t *dst; |
145 |
uint8_t *src; |
uint8_t *src; |
146 |
|
|
|
|
|
147 |
dst = image->y - (EDGE_SIZE + EDGE_SIZE * edged_width); |
dst = image->y - (EDGE_SIZE + EDGE_SIZE * edged_width); |
148 |
src = image->y; |
src = image->y; |
149 |
|
|
150 |
|
/* According to the Standard Clause 7.6.4, padding is done starting at 16 |
151 |
|
* pixel width and height multiples. This was not respected in old xvids */ |
152 |
|
if (bs_version >= SETEDGES_BUG_BEFORE) { |
153 |
|
width = (width+15)&~15; |
154 |
|
height = (height+15)&~15; |
155 |
|
} |
156 |
|
|
157 |
|
width2 = width/2; |
158 |
|
|
159 |
for (i = 0; i < EDGE_SIZE; i++) { |
for (i = 0; i < EDGE_SIZE; i++) { |
160 |
memset(dst, *src, EDGE_SIZE); |
memset(dst, *src, EDGE_SIZE); |
161 |
memcpy(dst + EDGE_SIZE, src, width); |
memcpy(dst + EDGE_SIZE, src, width); |
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, |
248 |
uint32_t rounding) |
uint32_t rounding) |
249 |
{ |
{ |
250 |
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 */ |
251 |
const uint32_t stride_add = 7 * edged_width; |
const uint32_t stride_add = 7 * edged_width; |
252 |
|
|
253 |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
uint8_t *n_ptr; |
254 |
|
uint8_t *h_ptr, *v_ptr, *hv_ptr; |
255 |
uint32_t x, y; |
uint32_t x, y; |
256 |
|
|
257 |
|
n_ptr = (uint8_t*)refn; |
258 |
n_ptr = refn->y; |
h_ptr = refh; |
259 |
h_ptr = refh->y; |
v_ptr = refv; |
|
v_ptr = refv->y; |
|
|
hv_ptr = refhv->y; |
|
260 |
|
|
261 |
n_ptr -= offset; |
n_ptr -= offset; |
262 |
h_ptr -= offset; |
h_ptr -= offset; |
263 |
v_ptr -= offset; |
v_ptr -= offset; |
264 |
|
|
265 |
|
/* Note we initialize the hv pointer later, as we can optimize code a bit |
266 |
|
* doing it down to up in quarterpel and up to down in halfpel */ |
267 |
|
if(quarterpel) { |
268 |
|
|
269 |
|
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
270 |
|
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
271 |
|
interpolate8x8_6tap_lowpass_h(h_ptr, n_ptr, edged_width, rounding); |
272 |
|
interpolate8x8_6tap_lowpass_v(v_ptr, n_ptr, edged_width, rounding); |
273 |
|
|
274 |
|
n_ptr += 8; |
275 |
|
h_ptr += 8; |
276 |
|
v_ptr += 8; |
277 |
|
} |
278 |
|
|
279 |
|
n_ptr += EDGE_SIZE; |
280 |
|
h_ptr += EDGE_SIZE; |
281 |
|
v_ptr += EDGE_SIZE; |
282 |
|
|
283 |
|
h_ptr += stride_add; |
284 |
|
v_ptr += stride_add; |
285 |
|
n_ptr += stride_add; |
286 |
|
} |
287 |
|
|
288 |
|
h_ptr = refh + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; |
289 |
|
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) { |
292 |
|
hv_ptr -= stride_add; |
293 |
|
h_ptr -= stride_add; |
294 |
|
hv_ptr -= EDGE_SIZE; |
295 |
|
h_ptr -= EDGE_SIZE; |
296 |
|
|
297 |
|
for (x = 0; x < (edged_width - EDGE_SIZE); x = x + 8) { |
298 |
|
hv_ptr -= 8; |
299 |
|
h_ptr -= 8; |
300 |
|
interpolate8x8_6tap_lowpass_v(hv_ptr, h_ptr, edged_width, rounding); |
301 |
|
} |
302 |
|
} |
303 |
|
} else { |
304 |
|
|
305 |
|
hv_ptr = refhv; |
306 |
hv_ptr -= offset; |
hv_ptr -= offset; |
307 |
|
|
308 |
for (y = 0; y < edged_height; y = y + 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
309 |
for (x = 0; x < edged_width; x = x + 8) { |
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
310 |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); |
311 |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); |
312 |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); |
316 |
v_ptr += 8; |
v_ptr += 8; |
317 |
hv_ptr += 8; |
hv_ptr += 8; |
318 |
} |
} |
319 |
|
|
320 |
|
h_ptr += EDGE_SIZE; |
321 |
|
v_ptr += EDGE_SIZE; |
322 |
|
hv_ptr += EDGE_SIZE; |
323 |
|
n_ptr += EDGE_SIZE; |
324 |
|
|
325 |
h_ptr += stride_add; |
h_ptr += stride_add; |
326 |
v_ptr += stride_add; |
v_ptr += stride_add; |
327 |
hv_ptr += stride_add; |
hv_ptr += stride_add; |
328 |
n_ptr += stride_add; |
n_ptr += stride_add; |
329 |
} |
} |
330 |
|
} |
331 |
|
} |
332 |
|
|
333 |
|
|
334 |
/* |
/* |
335 |
interpolate_halfpel_h( |
chroma optimize filter, invented by mf |
336 |
refh->y - offset, |
a chroma pixel is average from the surrounding pixels, when the |
337 |
refn->y - offset, |
correpsonding luma pixels are pure black or white. |
|
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); |
|
338 |
*/ |
*/ |
339 |
|
|
340 |
/* uv-image-based compensation |
void |
341 |
offset = EDGE_SIZE2 * (edged_width / 2 + 1); |
image_chroma_optimize(IMAGE * img, int width, int height, int edged_width) |
342 |
|
{ |
343 |
|
int x,y; |
344 |
|
int pixels = 0; |
345 |
|
|
346 |
|
for (y = 1; y < height/2 - 1; y++) |
347 |
|
for (x = 1; x < width/2 - 1; x++) |
348 |
|
{ |
349 |
|
#define IS_PURE(a) ((a)<=16||(a)>=235) |
350 |
|
#define IMG_Y(Y,X) img->y[(Y)*edged_width + (X)] |
351 |
|
#define IMG_U(Y,X) img->u[(Y)*edged_width/2 + (X)] |
352 |
|
#define IMG_V(Y,X) img->v[(Y)*edged_width/2 + (X)] |
353 |
|
|
354 |
|
if (IS_PURE(IMG_Y(y*2 ,x*2 )) && |
355 |
|
IS_PURE(IMG_Y(y*2 ,x*2+1)) && |
356 |
|
IS_PURE(IMG_Y(y*2+1,x*2 )) && |
357 |
|
IS_PURE(IMG_Y(y*2+1,x*2+1))) |
358 |
|
{ |
359 |
|
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; |
360 |
|
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; |
361 |
|
pixels++; |
362 |
|
} |
363 |
|
|
364 |
|
#undef IS_PURE |
365 |
|
#undef IMG_Y |
366 |
|
#undef IMG_U |
367 |
|
#undef IMG_V |
368 |
|
} |
369 |
|
|
370 |
|
DPRINTF(XVID_DEBUG_DEBUG,"chroma_optimized_pixels = %i/%i\n", pixels, width*height/4); |
371 |
|
} |
372 |
|
|
373 |
|
|
374 |
|
|
375 |
|
|
376 |
|
|
377 |
interpolate_halfpel_h( |
/* |
378 |
refh->u - offset, |
perform safe packed colorspace conversion, by splitting |
379 |
refn->u - offset, |
the image up into an optimized area (pixel width divisible by 16), |
380 |
edged_width / 2, edged_height / 2, |
and two unoptimized/plain-c areas (pixel width divisible by 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); |
|
381 |
*/ |
*/ |
382 |
|
|
383 |
|
static void |
384 |
|
safe_packed_conv(uint8_t * x_ptr, int x_stride, |
385 |
|
uint8_t * y_ptr, uint8_t * u_ptr, uint8_t * v_ptr, |
386 |
|
int y_stride, int uv_stride, |
387 |
|
int width, int height, int vflip, |
388 |
|
packedFunc * func_opt, packedFunc func_c, |
389 |
|
int size, int interlacing) |
390 |
|
{ |
391 |
|
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) |
396 |
|
{ |
397 |
|
width_opt = width & (~15); |
398 |
|
width_c = (width - width_opt) & (~1); |
399 |
} |
} |
400 |
|
else if (func_opt != func_c && !(width&1) && (size==3)) |
401 |
|
{ |
402 |
|
/* 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; |
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, |
420 |
|
y_ptr, u_ptr, v_ptr, y_stride, uv_stride, |
421 |
|
width_opt, height_opt, vflip); |
422 |
|
|
423 |
|
if (width_c) |
424 |
|
{ |
425 |
|
func_c(x_ptr + size*width_opt, x_stride, |
426 |
|
y_ptr + width_opt, u_ptr + width_opt/2, v_ptr + width_opt/2, |
427 |
|
y_stride, uv_stride, width_c, height_opt, vflip); |
428 |
|
} |
429 |
|
} |
430 |
|
|
431 |
|
|
432 |
|
|
433 |
int |
int |
435 |
uint32_t width, |
uint32_t width, |
436 |
int height, |
int height, |
437 |
uint32_t edged_width, |
uint32_t edged_width, |
438 |
uint8_t * src, |
uint8_t * src[4], |
439 |
int csp) |
int src_stride[4], |
440 |
{ |
int csp, |
441 |
|
int interlacing) |
|
/* if (csp & XVID_CSP_VFLIP) |
|
442 |
{ |
{ |
443 |
height = -height; |
const int edged_width2 = edged_width/2; |
444 |
} |
const int width2 = width/2; |
445 |
*/ |
const int height2 = height/2; |
446 |
|
#if 0 |
447 |
|
const int height_signed = (csp & XVID_CSP_VFLIP) ? -height : height; |
448 |
|
#endif |
449 |
|
|
450 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
451 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
452 |
rgb555_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
453 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
454 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
455 |
|
interlacing?rgb555i_to_yv12 :rgb555_to_yv12, |
456 |
|
interlacing?rgb555i_to_yv12_c:rgb555_to_yv12_c, 2, interlacing); |
457 |
|
break; |
458 |
|
|
459 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
460 |
rgb565_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
461 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
462 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
463 |
|
interlacing?rgb565i_to_yv12 :rgb565_to_yv12, |
464 |
|
interlacing?rgb565i_to_yv12_c:rgb565_to_yv12_c, 2, interlacing); |
465 |
|
break; |
466 |
|
|
467 |
|
|
468 |
case XVID_CSP_RGB24: |
case XVID_CSP_BGR: |
469 |
rgb24_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
470 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
471 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
472 |
|
interlacing?bgri_to_yv12 :bgr_to_yv12, |
473 |
|
interlacing?bgri_to_yv12_c:bgr_to_yv12_c, 3, interlacing); |
474 |
|
break; |
475 |
|
|
476 |
case XVID_CSP_RGB32: |
case XVID_CSP_BGRA: |
477 |
rgb32_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
478 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
479 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
480 |
|
interlacing?bgrai_to_yv12 :bgra_to_yv12, |
481 |
|
interlacing?bgrai_to_yv12_c:bgra_to_yv12_c, 4, interlacing); |
482 |
|
break; |
483 |
|
|
484 |
case XVID_CSP_I420: |
case XVID_CSP_ABGR : |
485 |
yuv_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
486 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
487 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
488 |
|
interlacing?abgri_to_yv12 :abgr_to_yv12, |
489 |
|
interlacing?abgri_to_yv12_c:abgr_to_yv12_c, 4, interlacing); |
490 |
|
break; |
491 |
|
|
492 |
case XVID_CSP_YV12: /* u/v swapped */ |
case XVID_CSP_RGB: |
493 |
yuv_to_yv12(image->y, image->v, image->u, src, width, height, |
safe_packed_conv( |
494 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
495 |
return 0; |
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; |
499 |
|
|
500 |
|
case XVID_CSP_RGBA : |
501 |
|
safe_packed_conv( |
502 |
|
src[0], src_stride[0], image->y, image->u, image->v, |
503 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
504 |
|
interlacing?rgbai_to_yv12 :rgba_to_yv12, |
505 |
|
interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4, interlacing); |
506 |
|
break; |
507 |
|
|
508 |
|
case XVID_CSP_ARGB: |
509 |
|
safe_packed_conv( |
510 |
|
src[0], src_stride[0], image->y, image->u, image->v, |
511 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
512 |
|
interlacing?argbi_to_yv12 : argb_to_yv12, |
513 |
|
interlacing?argbi_to_yv12_c: argb_to_yv12_c, 4, interlacing); |
514 |
|
break; |
515 |
|
|
516 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
517 |
yuyv_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
518 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
519 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
520 |
|
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
521 |
|
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2, interlacing); |
522 |
|
break; |
523 |
|
|
524 |
case XVID_CSP_YVYU: /* u/v swapped */ |
case XVID_CSP_YVYU: /* u/v swapped */ |
525 |
yuyv_to_yv12(image->y, image->v, image->u, src, width, height, |
safe_packed_conv( |
526 |
edged_width); |
src[0], src_stride[0], image->y, image->v, image->u, |
527 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
528 |
|
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
529 |
|
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2, interlacing); |
530 |
|
break; |
531 |
|
|
532 |
case XVID_CSP_UYVY: |
case XVID_CSP_UYVY: |
533 |
uyvy_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
534 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
535 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
536 |
|
interlacing?uyvyi_to_yv12 :uyvy_to_yv12, |
537 |
|
interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2, interlacing); |
538 |
|
break; |
539 |
|
|
540 |
case XVID_CSP_USER: |
case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
541 |
user_to_yuv_c(image->y, image->u, image->v, edged_width, |
yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
542 |
(DEC_PICTURE *) src, width, height); |
src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
543 |
return 0; |
src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); |
544 |
|
break; |
545 |
|
|
546 |
|
case XVID_CSP_YV12: /* YCrCb == YVA == U and V plane swapped */ |
547 |
|
yv12_to_yv12(image->y, image->v, image->u, edged_width, edged_width2, |
548 |
|
src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
549 |
|
src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); |
550 |
|
break; |
551 |
|
|
552 |
|
case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ |
553 |
|
yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
554 |
|
src[0], src[1], src[2], src_stride[0], src_stride[1], /* v: dst_stride[2] not yet supported */ |
555 |
|
width, height, (csp & XVID_CSP_VFLIP)); |
556 |
|
break; |
557 |
|
|
558 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
559 |
break; |
break; |
560 |
|
|
561 |
|
default : |
562 |
|
return -1; |
563 |
} |
} |
564 |
|
|
565 |
return -1; |
|
566 |
|
/* pad out image when the width and/or height is not a multiple of 16 */ |
567 |
|
|
568 |
|
if (width & 15) |
569 |
|
{ |
570 |
|
int i; |
571 |
|
int pad_width = 16 - (width&15); |
572 |
|
for (i = 0; i < height; i++) |
573 |
|
{ |
574 |
|
memset(image->y + i*edged_width + width, |
575 |
|
*(image->y + i*edged_width + width - 1), pad_width); |
576 |
|
} |
577 |
|
for (i = 0; i < height/2; i++) |
578 |
|
{ |
579 |
|
memset(image->u + i*edged_width2 + width2, |
580 |
|
*(image->u + i*edged_width2 + width2 - 1),pad_width/2); |
581 |
|
memset(image->v + i*edged_width2 + width2, |
582 |
|
*(image->v + i*edged_width2 + width2 - 1),pad_width/2); |
583 |
|
} |
584 |
|
} |
585 |
|
|
586 |
|
if (height & 15) |
587 |
|
{ |
588 |
|
int pad_height = 16 - (height&15); |
589 |
|
int length = ((width+15)/16)*16; |
590 |
|
int i; |
591 |
|
for (i = 0; i < pad_height; i++) |
592 |
|
{ |
593 |
|
memcpy(image->y + (height+i)*edged_width, |
594 |
|
image->y + (height-1)*edged_width,length); |
595 |
|
} |
596 |
|
|
597 |
|
for (i = 0; i < pad_height/2; i++) |
598 |
|
{ |
599 |
|
memcpy(image->u + (height2+i)*edged_width2, |
600 |
|
image->u + (height2-1)*edged_width2,length/2); |
601 |
|
memcpy(image->v + (height2+i)*edged_width2, |
602 |
|
image->v + (height2-1)*edged_width2,length/2); |
603 |
|
} |
604 |
|
} |
605 |
|
|
606 |
|
/* |
607 |
|
if (interlacing) |
608 |
|
image_printf(image, edged_width, height, 5,5, "[i]"); |
609 |
|
image_dump_yuvpgm(image, edged_width, ((width+15)/16)*16, ((height+15)/16)*16, "\\encode.pgm"); |
610 |
|
*/ |
611 |
|
return 0; |
612 |
} |
} |
613 |
|
|
614 |
|
|
618 |
uint32_t width, |
uint32_t width, |
619 |
int height, |
int height, |
620 |
uint32_t edged_width, |
uint32_t edged_width, |
621 |
uint8_t * dst, |
uint8_t * dst[4], |
622 |
uint32_t dst_stride, |
int dst_stride[4], |
623 |
int csp) |
int csp, |
624 |
|
int interlacing) |
625 |
{ |
{ |
626 |
if (csp & XVID_CSP_VFLIP) { |
const int edged_width2 = edged_width/2; |
627 |
height = -height; |
int height2 = height/2; |
628 |
} |
|
629 |
|
/* |
630 |
|
if (interlacing) |
631 |
|
image_printf(image, edged_width, height, 5,100, "[i]=%i,%i",width,height); |
632 |
|
image_dump_yuvpgm(image, edged_width, width, height, "\\decode.pgm"); |
633 |
|
*/ |
634 |
|
|
635 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
636 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
637 |
yv12_to_rgb555(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
638 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->u, image->v, |
639 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
640 |
|
interlacing?yv12_to_rgb555i :yv12_to_rgb555, |
641 |
|
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: |
645 |
yv12_to_rgb565(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
646 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->u, image->v, |
647 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
648 |
|
interlacing?yv12_to_rgb565i :yv12_to_rgb565, |
649 |
case XVID_CSP_RGB24: |
interlacing?yv12_to_rgb565i_c:yv12_to_rgb565_c, 2, interlacing); |
650 |
yv12_to_rgb24(dst, dst_stride, image->y, image->u, image->v, |
return 0; |
651 |
edged_width, edged_width / 2, width, height); |
|
652 |
return 0; |
case XVID_CSP_BGR: |
653 |
|
safe_packed_conv( |
654 |
case XVID_CSP_RGB32: |
dst[0], dst_stride[0], image->y, image->u, image->v, |
655 |
yv12_to_rgb32(dst, dst_stride, image->y, image->u, image->v, |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
656 |
edged_width, edged_width / 2, width, height); |
interlacing?yv12_to_bgri :yv12_to_bgr, |
657 |
return 0; |
interlacing?yv12_to_bgri_c:yv12_to_bgr_c, 3, interlacing); |
658 |
|
return 0; |
659 |
case XVID_CSP_I420: |
|
660 |
yv12_to_yuv(dst, dst_stride, image->y, image->u, image->v, edged_width, |
case XVID_CSP_BGRA: |
661 |
edged_width / 2, width, height); |
safe_packed_conv( |
662 |
return 0; |
dst[0], dst_stride[0], image->y, image->u, image->v, |
663 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
664 |
case XVID_CSP_YV12: /* u,v swapped */ |
interlacing?yv12_to_bgrai :yv12_to_bgra, |
665 |
yv12_to_yuv(dst, dst_stride, image->y, image->v, image->u, edged_width, |
interlacing?yv12_to_bgrai_c:yv12_to_bgra_c, 4, interlacing); |
666 |
edged_width / 2, width, height); |
return 0; |
667 |
|
|
668 |
|
case XVID_CSP_ABGR: |
669 |
|
safe_packed_conv( |
670 |
|
dst[0], dst_stride[0], image->y, image->u, image->v, |
671 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
672 |
|
interlacing?yv12_to_abgri :yv12_to_abgr, |
673 |
|
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; |
683 |
|
|
684 |
|
case XVID_CSP_RGBA: |
685 |
|
safe_packed_conv( |
686 |
|
dst[0], dst_stride[0], image->y, image->u, image->v, |
687 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
688 |
|
interlacing?yv12_to_rgbai :yv12_to_rgba, |
689 |
|
interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4, interlacing); |
690 |
|
return 0; |
691 |
|
|
692 |
|
case XVID_CSP_ARGB: |
693 |
|
safe_packed_conv( |
694 |
|
dst[0], dst_stride[0], image->y, image->u, image->v, |
695 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
696 |
|
interlacing?yv12_to_argbi :yv12_to_argb, |
697 |
|
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: |
701 |
yv12_to_yuyv(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
702 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->u, image->v, |
703 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
704 |
|
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
705 |
|
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 */ |
709 |
yv12_to_yuyv(dst, dst_stride, image->y, image->v, image->u, |
safe_packed_conv( |
710 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->v, image->u, |
711 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
712 |
|
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
713 |
|
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: |
717 |
yv12_to_uyvy(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
718 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->u, image->v, |
719 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
720 |
|
interlacing?yv12_to_uyvyi :yv12_to_uyvy, |
721 |
case XVID_CSP_USER: |
interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2, interlacing); |
722 |
((DEC_PICTURE *) dst)->y = image->y; |
return 0; |
723 |
((DEC_PICTURE *) dst)->u = image->u; |
|
724 |
((DEC_PICTURE *) dst)->v = image->v; |
case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
725 |
((DEC_PICTURE *) dst)->stride_y = edged_width; |
yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, |
726 |
((DEC_PICTURE *) dst)->stride_uv = edged_width / 2; |
dst_stride[0], dst_stride[0]/2, |
727 |
|
image->y, image->u, image->v, edged_width, edged_width2, |
728 |
|
width, height, (csp & XVID_CSP_VFLIP)); |
729 |
|
return 0; |
730 |
|
|
731 |
|
case XVID_CSP_YV12: /* YCrCb == YVU == U and V plane swapped */ |
732 |
|
yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, |
733 |
|
dst_stride[0], dst_stride[0]/2, |
734 |
|
image->y, image->v, image->u, edged_width, edged_width2, |
735 |
|
width, height, (csp & XVID_CSP_VFLIP)); |
736 |
|
return 0; |
737 |
|
|
738 |
|
case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ |
739 |
|
yv12_to_yv12(dst[0], dst[1], dst[2], |
740 |
|
dst_stride[0], dst_stride[1], /* v: dst_stride[2] not yet supported */ |
741 |
|
image->y, image->u, image->v, edged_width, edged_width2, |
742 |
|
width, height, (csp & XVID_CSP_VFLIP)); |
743 |
|
return 0; |
744 |
|
|
745 |
|
case XVID_CSP_INTERNAL : |
746 |
|
dst[0] = image->y; |
747 |
|
dst[1] = image->u; |
748 |
|
dst[2] = image->v; |
749 |
|
dst_stride[0] = edged_width; |
750 |
|
dst_stride[1] = edged_width/2; |
751 |
|
dst_stride[2] = edged_width/2; |
752 |
return 0; |
return 0; |
753 |
|
|
754 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
755 |
case XVID_CSP_EXTERN: |
case XVID_CSP_SLICE: |
756 |
return 0; |
return 0; |
757 |
|
|
758 |
} |
} |
792 |
return psnr_y; |
return psnr_y; |
793 |
} |
} |
794 |
|
|
795 |
|
|
796 |
|
float sse_to_PSNR(long sse, int pixels) |
797 |
|
{ |
798 |
|
if (sse==0) |
799 |
|
return 99.99F; |
800 |
|
|
801 |
|
return 48.131F - 10*(float)log10((float)sse/(float)(pixels)); /* log10(255*255)=4.8131 */ |
802 |
|
|
803 |
|
} |
804 |
|
|
805 |
|
long plane_sse(uint8_t *orig, |
806 |
|
uint8_t *recon, |
807 |
|
uint16_t stride, |
808 |
|
uint16_t width, |
809 |
|
uint16_t height) |
810 |
|
{ |
811 |
|
int y, bwidth, bheight; |
812 |
|
long sse = 0; |
813 |
|
|
814 |
|
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++) { |
854 |
|
int diff; |
855 |
|
diff = *(orig + x) - *(recon + x); |
856 |
|
sse += diff * diff; |
857 |
|
} |
858 |
|
orig += stride; |
859 |
|
recon += stride; |
860 |
|
} |
861 |
|
|
862 |
|
return (sse); |
863 |
|
} |
864 |
|
|
865 |
|
void image_block_variance(IMAGE * orig_image, |
866 |
|
uint16_t stride, |
867 |
|
MACROBLOCK *mbs, |
868 |
|
uint16_t mb_width, |
869 |
|
uint16_t mb_height) |
870 |
|
{ |
871 |
|
DECLARE_ALIGNED_MATRIX(sums, 1, 4, uint16_t, CACHE_LINE); |
872 |
|
DECLARE_ALIGNED_MATRIX(squares, 1, 4, uint32_t, CACHE_LINE); |
873 |
|
|
874 |
|
int x, y, i, j; |
875 |
|
uint8_t *orig_y = orig_image->y; |
876 |
|
uint8_t *orig_u = orig_image->u; |
877 |
|
uint8_t *orig_v = orig_image->v; |
878 |
|
|
879 |
|
for (y = 0; y < mb_height; y++) { |
880 |
|
for (x = 0; x < mb_width; x++) { |
881 |
|
MACROBLOCK *pMB = &mbs[x + y * mb_width]; |
882 |
|
uint32_t var4[4]; |
883 |
|
uint32_t sum = 0, square = 0; |
884 |
|
|
885 |
|
/* y-blocks */ |
886 |
|
for (j = 0; j < 2; j++) { |
887 |
|
for (i = 0; i < 2; i++) { |
888 |
|
int lsum = blocksum8(orig_y + ((y<<4) + (j<<3))*stride + (x<<4) + (i<<3), |
889 |
|
stride, sums, squares); |
890 |
|
int lsquare = (squares[0] + squares[1] + squares[2] + squares[3])<<6; |
891 |
|
|
892 |
|
sum += lsum; |
893 |
|
square += lsquare; |
894 |
|
|
895 |
|
var4[0] = (squares[0]<<4) - sums[0]*sums[0]; |
896 |
|
var4[1] = (squares[1]<<4) - sums[1]*sums[1]; |
897 |
|
var4[2] = (squares[2]<<4) - sums[2]*sums[2]; |
898 |
|
var4[3] = (squares[3]<<4) - sums[3]*sums[3]; |
899 |
|
|
900 |
|
pMB->rel_var8[j*2 + i] = lsquare - lsum*lsum; |
901 |
|
if (pMB->rel_var8[j*2 + i]) |
902 |
|
pMB->rel_var8[j*2 + i] = ((var4[0] + var4[1] + var4[2] + var4[3])<<8) / |
903 |
|
pMB->rel_var8[j*2 + i]; /* 4*(Var(Di)/Var(D)) */ |
904 |
|
else |
905 |
|
pMB->rel_var8[j*2 + i] = 64; |
906 |
|
} |
907 |
|
} |
908 |
|
|
909 |
|
/* u */ |
910 |
|
{ |
911 |
|
int lsum = blocksum8(orig_u + (y<<3)*(stride>>1) + (x<<3), |
912 |
|
stride, sums, squares); |
913 |
|
int lsquare = (squares[0] + squares[1] + squares[2] + squares[3])<<6; |
914 |
|
|
915 |
|
sum += lsum; |
916 |
|
square += lsquare; |
917 |
|
|
918 |
|
var4[0] = (squares[0]<<4) - sums[0]*sums[0]; |
919 |
|
var4[1] = (squares[1]<<4) - sums[1]*sums[1]; |
920 |
|
var4[2] = (squares[2]<<4) - sums[2]*sums[2]; |
921 |
|
var4[3] = (squares[3]<<4) - sums[3]*sums[3]; |
922 |
|
|
923 |
|
pMB->rel_var8[4] = lsquare - lsum*lsum; |
924 |
|
if (pMB->rel_var8[4]) |
925 |
|
pMB->rel_var8[4] = ((var4[0] + var4[1] + var4[2] + var4[3])<<8) / |
926 |
|
pMB->rel_var8[4]; /* 4*(Var(Di)/Var(D)) */ |
927 |
|
else |
928 |
|
pMB->rel_var8[4] = 64; |
929 |
|
} |
930 |
|
|
931 |
|
/* v */ |
932 |
|
{ |
933 |
|
int lsum = blocksum8(orig_v + (y<<3)*(stride>>1) + (x<<3), |
934 |
|
stride, sums, squares); |
935 |
|
int lsquare = (squares[0] + squares[1] + squares[2] + squares[3])<<6; |
936 |
|
|
937 |
|
sum += lsum; |
938 |
|
square += lsquare; |
939 |
|
|
940 |
|
var4[0] = (squares[0]<<4) - sums[0]*sums[0]; |
941 |
|
var4[1] = (squares[1]<<4) - sums[1]*sums[1]; |
942 |
|
var4[2] = (squares[2]<<4) - sums[2]*sums[2]; |
943 |
|
var4[3] = (squares[3]<<4) - sums[3]*sums[3]; |
944 |
|
|
945 |
|
pMB->rel_var8[5] = lsquare - lsum*lsum; |
946 |
|
if (pMB->rel_var8[5]) |
947 |
|
pMB->rel_var8[5] = ((var4[0] + var4[1] + var4[2] + var4[3])<<8) / |
948 |
|
pMB->rel_var8[5]; /* 4*(Var(Di)/Var(D)) */ |
949 |
|
else |
950 |
|
pMB->rel_var8[5] = 64; |
951 |
|
} |
952 |
|
|
953 |
|
} |
954 |
|
} |
955 |
|
} |
956 |
|
|
957 |
#if 0 |
#if 0 |
958 |
|
|
959 |
#include <stdio.h> |
#include <stdio.h> |
1050 |
} |
} |
1051 |
|
|
1052 |
|
|
|
#define ABS(X) (((X)>0)?(X):-(X)) |
|
1053 |
float |
float |
1054 |
image_mad(const IMAGE * img1, |
image_mad(const IMAGE * img1, |
1055 |
const IMAGE * img2, |
const IMAGE * img2, |
1066 |
|
|
1067 |
for (y = 0; y < height; y++) |
for (y = 0; y < height; y++) |
1068 |
for (x = 0; x < width; x++) |
for (x = 0; x < width; x++) |
1069 |
sum += ABS(img1->y[x + y * stride] - img2->y[x + y * stride]); |
sum += abs(img1->y[x + y * stride] - img2->y[x + y * stride]); |
1070 |
|
|
1071 |
for (y = 0; y < height2; y++) |
for (y = 0; y < height2; y++) |
1072 |
for (x = 0; x < width2; x++) |
for (x = 0; x < width2; x++) |
1073 |
sum += ABS(img1->u[x + y * stride2] - img2->u[x + y * stride2]); |
sum += abs(img1->u[x + y * stride2] - img2->u[x + y * stride2]); |
1074 |
|
|
1075 |
for (y = 0; y < height2; y++) |
for (y = 0; y < height2; y++) |
1076 |
for (x = 0; x < width2; x++) |
for (x = 0; x < width2; x++) |
1077 |
sum += ABS(img1->v[x + y * stride2] - img2->v[x + y * stride2]); |
sum += abs(img1->v[x + y * stride2] - img2->v[x + y * stride2]); |
1078 |
|
|
1079 |
return (float) sum / (width * height * 3 / 2); |
return (float) sum / (width * height * 3 / 2); |
1080 |
} |
} |
1081 |
|
|
1082 |
void |
void |
1083 |
output_slice(IMAGE * cur, int std, int width, XVID_DEC_PICTURE* 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) { |
1084 |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
1085 |
int std2 = std >> 1; |
int stride2 = stride >> 1; |
1086 |
int w = mbl << 4, w2,i; |
int w = mbl << 4, w2,i; |
1087 |
|
|
1088 |
if(w > width) |
if(w > width) |
1089 |
w = width; |
w = width; |
1090 |
w2 = w >> 1; |
w2 = w >> 1; |
1091 |
|
|
1092 |
dY = (uint8_t*)out_frm->y + (mby << 4) * out_frm->stride_y + (mbx << 4); |
dY = (uint8_t*)out_frm->plane[0] + (mby << 4) * out_frm->stride[0] + (mbx << 4); |
1093 |
dU = (uint8_t*)out_frm->u + (mby << 3) * out_frm->stride_u + (mbx << 3); |
dU = (uint8_t*)out_frm->plane[1] + (mby << 3) * out_frm->stride[1] + (mbx << 3); |
1094 |
dV = (uint8_t*)out_frm->v + (mby << 3) * out_frm->stride_v + (mbx << 3); |
dV = (uint8_t*)out_frm->plane[2] + (mby << 3) * out_frm->stride[2] + (mbx << 3); |
1095 |
sY = cur->y + (mby << 4) * std + (mbx << 4); |
sY = cur->y + (mby << 4) * stride + (mbx << 4); |
1096 |
sU = cur->u + (mby << 3) * std2 + (mbx << 3); |
sU = cur->u + (mby << 3) * stride2 + (mbx << 3); |
1097 |
sV = cur->v + (mby << 3) * std2 + (mbx << 3); |
sV = cur->v + (mby << 3) * stride2 + (mbx << 3); |
1098 |
|
|
1099 |
for(i = 0 ; i < 16 ; i++) { |
for(i = 0 ; i < 16 ; i++) { |
1100 |
memcpy(dY,sY,w); |
memcpy(dY,sY,w); |
1101 |
dY += out_frm->stride_y; |
dY += out_frm->stride[0]; |
1102 |
sY += std; |
sY += stride; |
1103 |
} |
} |
1104 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
1105 |
memcpy(dU,sU,w2); |
memcpy(dU,sU,w2); |
1106 |
dU += out_frm->stride_u; |
dU += out_frm->stride[1]; |
1107 |
sU += std2; |
sU += stride2; |
1108 |
} |
} |
1109 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
1110 |
memcpy(dV,sV,w2); |
memcpy(dV,sV,w2); |
1111 |
dV += out_frm->stride_v; |
dV += out_frm->stride[2]; |
1112 |
sV += std2; |
sV += stride2; |
1113 |
|
} |
1114 |
|
} |
1115 |
|
|
1116 |
|
|
1117 |
|
void |
1118 |
|
image_clear(IMAGE * img, int width, int height, int edged_width, |
1119 |
|
int y, int u, int v) |
1120 |
|
{ |
1121 |
|
uint8_t * p; |
1122 |
|
int i; |
1123 |
|
|
1124 |
|
p = img->y; |
1125 |
|
for (i = 0; i < height; i++) { |
1126 |
|
memset(p, y, width); |
1127 |
|
p += edged_width; |
1128 |
|
} |
1129 |
|
|
1130 |
|
p = img->u; |
1131 |
|
for (i = 0; i < height/2; i++) { |
1132 |
|
memset(p, u, width/2); |
1133 |
|
p += edged_width/2; |
1134 |
|
} |
1135 |
|
|
1136 |
|
p = img->v; |
1137 |
|
for (i = 0; i < height/2; i++) { |
1138 |
|
memset(p, v, width/2); |
1139 |
|
p += edged_width/2; |
1140 |
|
} |
1141 |
|
} |
1142 |
|
|
1143 |
|
/****************************************************************************/ |
1144 |
|
|
1145 |
|
static void (*deintl_core)(uint8_t *, int width, int height, const int stride) = 0; |
1146 |
|
extern void xvid_deinterlace_sse(uint8_t *, int width, int height, const int stride); |
1147 |
|
|
1148 |
|
#define CLIP_255(x) ( ((x)&~255) ? ((-(x)) >> (8*sizeof((x))-1))&0xff : (x) ) |
1149 |
|
|
1150 |
|
static void deinterlace_c(uint8_t *pix, int width, int height, const int bps) |
1151 |
|
{ |
1152 |
|
pix += bps; |
1153 |
|
while(width-->0) |
1154 |
|
{ |
1155 |
|
int p1 = pix[-bps]; |
1156 |
|
int p2 = pix[0]; |
1157 |
|
int p0 = p2; |
1158 |
|
int j = (height>>1) - 1; |
1159 |
|
int V; |
1160 |
|
unsigned char *P = pix++; |
1161 |
|
while(j-->0) |
1162 |
|
{ |
1163 |
|
const int p3 = P[ bps]; |
1164 |
|
const int p4 = P[2*bps]; |
1165 |
|
V = ((p1+p3+1)>>1) + ((p2 - ((p0+p4+1)>>1)) >> 2); |
1166 |
|
P[0] = CLIP_255( V ); |
1167 |
|
p0 = p2; |
1168 |
|
p1 = p3; |
1169 |
|
p2 = p4; |
1170 |
|
P += 2*bps; |
1171 |
|
} |
1172 |
|
V = ((p1+p1+1)>>1) + ((p2 - ((p0+p2+1)>>1)) >> 2); |
1173 |
|
P[0] = CLIP_255( V ); |
1174 |
} |
} |
1175 |
} |
} |
1176 |
|
#undef CLIP_255 |
1177 |
|
|
1178 |
|
int xvid_image_deinterlace(xvid_image_t* img, int width, int height, int bottom_first) |
1179 |
|
{ |
1180 |
|
if (height&1) |
1181 |
|
return 0; |
1182 |
|
if (img->csp!=XVID_CSP_PLANAR && img->csp!=XVID_CSP_I420 && img->csp!=XVID_CSP_YV12) |
1183 |
|
return 0; /* not yet supported */ |
1184 |
|
if (deintl_core==0) { |
1185 |
|
deintl_core = deinterlace_c; |
1186 |
|
#if defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) |
1187 |
|
{ |
1188 |
|
int cpu_flags = check_cpu_features(); |
1189 |
|
if (cpu_flags & XVID_CPU_MMX) |
1190 |
|
deintl_core = xvid_deinterlace_sse; |
1191 |
|
} |
1192 |
|
#endif |
1193 |
|
} |
1194 |
|
if (!bottom_first) { |
1195 |
|
deintl_core(img->plane[0], width, height, img->stride[0]); |
1196 |
|
deintl_core(img->plane[1], width>>1, height>>1, img->stride[1]); |
1197 |
|
deintl_core(img->plane[2], width>>1, height>>1, img->stride[2]); |
1198 |
|
} |
1199 |
|
else { |
1200 |
|
deintl_core((uint8_t *)img->plane[0] + ( height -1)*img->stride[0], width, height, -img->stride[0]); |
1201 |
|
deintl_core((uint8_t *)img->plane[1] + ((height>>1)-1)*img->stride[1], width>>1, height>>1, -img->stride[1]); |
1202 |
|
deintl_core((uint8_t *)img->plane[2] + ((height>>1)-1)*img->stride[2], width>>1, height>>1, -img->stride[2]); |
1203 |
|
} |
1204 |
|
emms(); |
1205 |
|
|
1206 |
|
return 1; |
1207 |
|
} |
1208 |
|
|