1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* image stuff |
* - image module - |
5 |
|
* |
6 |
|
* Copyright(C) 2002 Peter Ross <pross@xvid.org> |
7 |
* |
* |
8 |
* This program is an implementation of a part of one or more MPEG-4 |
* This program is an implementation of a part of one or more MPEG-4 |
9 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
26 |
* |
* |
27 |
* You should have received a copy of the GNU General Public License |
* You should have received a copy of the GNU General Public License |
28 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
29 |
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
|
* |
|
|
*************************************************************************/ |
|
|
|
|
|
/************************************************************************** |
|
|
* |
|
|
* History: |
|
30 |
* |
* |
31 |
* 05.10.2002 support for interpolated images in qpel mode - Isibaar |
****************************************************************************/ |
|
* 01.05.2002 BFRAME image-based u,v interpolation |
|
|
* 22.04.2002 added some B-frame support |
|
|
* 14.04.2002 added image_dump_yuvpgm(), added image_mad() |
|
|
* XVID_CSP_USER input support |
|
|
* 09.04.2002 PSNR calculations - Isibaar |
|
|
* 06.04.2002 removed interlaced edging from U,V blocks (as per spec) |
|
|
* 26.03.2002 interlacing support (field-based edging in set_edges) |
|
|
* 26.01.2002 rgb555, rgb565 |
|
|
* 07.01.2001 commented u,v interpolation (not required for uv-block-based) |
|
|
* 23.12.2001 removed #ifdefs, added function pointers + init_common() |
|
|
* 22.12.2001 cpu #ifdefs |
|
|
* 19.12.2001 image_dump(); useful for debugging |
|
|
* 6.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
|
* |
|
|
*************************************************************************/ |
|
32 |
|
|
33 |
#include <stdlib.h> |
#include <stdlib.h> |
34 |
#include <string.h> // memcpy, memset |
#include <string.h> // memcpy, memset |
35 |
#include <math.h> |
#include <math.h> |
36 |
|
|
37 |
#include "../portab.h" |
#include "../portab.h" |
|
#include "../global.h" // XVID_CSP_XXX's |
|
38 |
#include "../xvid.h" // XVID_CSP_XXX's |
#include "../xvid.h" // XVID_CSP_XXX's |
39 |
#include "image.h" |
#include "image.h" |
40 |
#include "colorspace.h" |
#include "colorspace.h" |
41 |
#include "interpolate8x8.h" |
#include "interpolate8x8.h" |
|
#include "reduced.h" |
|
42 |
#include "../divx4.h" |
#include "../divx4.h" |
43 |
#include "../utils/mem_align.h" |
#include "../utils/mem_align.h" |
44 |
|
|
|
#include "font.h" // XXX: remove later |
|
|
|
|
45 |
#define SAFETY 64 |
#define SAFETY 64 |
46 |
#define EDGE_SIZE2 (EDGE_SIZE/2) |
#define EDGE_SIZE2 (EDGE_SIZE/2) |
47 |
|
|
142 |
uint32_t edged_width, |
uint32_t edged_width, |
143 |
uint32_t edged_height, |
uint32_t edged_height, |
144 |
uint32_t width, |
uint32_t width, |
145 |
uint32_t height) |
uint32_t height, |
146 |
|
uint32_t interlacing) |
147 |
{ |
{ |
148 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
149 |
const uint32_t width2 = width / 2; |
const uint32_t width2 = width / 2; |
156 |
src = image->y; |
src = image->y; |
157 |
|
|
158 |
for (i = 0; i < EDGE_SIZE; i++) { |
for (i = 0; i < EDGE_SIZE; i++) { |
159 |
|
/* // if interlacing, edges contain top-most data from each field |
160 |
|
if (interlacing && (i & 1)) { |
161 |
|
memset(dst, *(src + edged_width), EDGE_SIZE); |
162 |
|
memcpy(dst + EDGE_SIZE, src + edged_width, width); |
163 |
|
memset(dst + edged_width - EDGE_SIZE, |
164 |
|
*(src + edged_width + width - 1), EDGE_SIZE); |
165 |
|
} else {*/ |
166 |
memset(dst, *src, EDGE_SIZE); |
memset(dst, *src, EDGE_SIZE); |
167 |
memcpy(dst + EDGE_SIZE, src, width); |
memcpy(dst + EDGE_SIZE, src, width); |
168 |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
169 |
EDGE_SIZE); |
EDGE_SIZE); |
170 |
|
/*}*/ |
171 |
dst += edged_width; |
dst += edged_width; |
172 |
} |
} |
173 |
|
|
180 |
|
|
181 |
src -= edged_width; |
src -= edged_width; |
182 |
for (i = 0; i < EDGE_SIZE; i++) { |
for (i = 0; i < EDGE_SIZE; i++) { |
183 |
|
/* // if interlacing, edges contain bottom-most data from each field |
184 |
|
if (interlacing && !(i & 1)) { |
185 |
|
memset(dst, *(src - edged_width), EDGE_SIZE); |
186 |
|
memcpy(dst + EDGE_SIZE, src - edged_width, width); |
187 |
|
memset(dst + edged_width - EDGE_SIZE, |
188 |
|
*(src - edged_width + width - 1), EDGE_SIZE); |
189 |
|
} else {*/ |
190 |
memset(dst, *src, EDGE_SIZE); |
memset(dst, *src, EDGE_SIZE); |
191 |
memcpy(dst + EDGE_SIZE, src, width); |
memcpy(dst + EDGE_SIZE, src, width); |
192 |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
193 |
EDGE_SIZE); |
EDGE_SIZE); |
194 |
|
/*}*/ |
195 |
dst += edged_width; |
dst += edged_width; |
196 |
} |
} |
197 |
|
|
260 |
IMAGE * refhv, |
IMAGE * refhv, |
261 |
uint32_t edged_width, |
uint32_t edged_width, |
262 |
uint32_t edged_height, |
uint32_t edged_height, |
|
uint32_t quarterpel, |
|
263 |
uint32_t rounding) |
uint32_t rounding) |
264 |
{ |
{ |
265 |
const uint32_t offset = EDGE_SIZE2 * (edged_width + 1); // we only interpolate half of the edge area |
const uint32_t offset = EDGE_SIZE * (edged_width + 1); |
266 |
const uint32_t stride_add = 7 * edged_width; |
const uint32_t stride_add = 7 * edged_width; |
267 |
/* |
|
|
#ifdef BFRAMES |
|
|
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 |
|
|
*/ |
|
268 |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
269 |
uint32_t x, y; |
uint32_t x, y; |
270 |
|
|
279 |
v_ptr -= offset; |
v_ptr -= offset; |
280 |
hv_ptr -= offset; |
hv_ptr -= offset; |
281 |
|
|
282 |
if(quarterpel) { |
for (y = 0; y < edged_height; y = y + 8) { |
283 |
|
for (x = 0; x < edged_width; x = x + 8) { |
|
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
|
|
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
|
|
interpolate8x8_6tap_lowpass_h(h_ptr, n_ptr, edged_width, rounding); |
|
|
interpolate8x8_6tap_lowpass_v(v_ptr, n_ptr, edged_width, rounding); |
|
|
|
|
|
n_ptr += 8; |
|
|
h_ptr += 8; |
|
|
v_ptr += 8; |
|
|
} |
|
|
|
|
|
n_ptr += EDGE_SIZE; |
|
|
h_ptr += EDGE_SIZE; |
|
|
v_ptr += EDGE_SIZE; |
|
|
|
|
|
h_ptr += stride_add; |
|
|
v_ptr += stride_add; |
|
|
n_ptr += stride_add; |
|
|
} |
|
|
|
|
|
h_ptr = refh->y; |
|
|
h_ptr -= offset; |
|
|
|
|
|
for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { |
|
|
for (x = 0; x < (edged_width - EDGE_SIZE); x = x + 8) { |
|
|
interpolate8x8_6tap_lowpass_v(hv_ptr, h_ptr, edged_width, rounding); |
|
|
hv_ptr += 8; |
|
|
h_ptr += 8; |
|
|
} |
|
|
|
|
|
hv_ptr += EDGE_SIZE; |
|
|
h_ptr += EDGE_SIZE; |
|
|
|
|
|
hv_ptr += stride_add; |
|
|
h_ptr += stride_add; |
|
|
} |
|
|
} |
|
|
else { |
|
|
|
|
|
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
|
|
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
|
284 |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); |
285 |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); |
286 |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); |
290 |
v_ptr += 8; |
v_ptr += 8; |
291 |
hv_ptr += 8; |
hv_ptr += 8; |
292 |
} |
} |
|
|
|
|
h_ptr += EDGE_SIZE; |
|
|
v_ptr += EDGE_SIZE; |
|
|
hv_ptr += EDGE_SIZE; |
|
|
n_ptr += EDGE_SIZE; |
|
|
|
|
293 |
h_ptr += stride_add; |
h_ptr += stride_add; |
294 |
v_ptr += stride_add; |
v_ptr += stride_add; |
295 |
hv_ptr += stride_add; |
hv_ptr += stride_add; |
296 |
n_ptr += stride_add; |
n_ptr += stride_add; |
297 |
} |
} |
|
} |
|
|
/* |
|
|
#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); |
|
298 |
|
|
|
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 |
|
|
*/ |
|
299 |
/* |
/* |
300 |
interpolate_halfpel_h( |
interpolate_halfpel_h( |
301 |
refh->y - offset, |
refh->y - offset, |
359 |
} |
} |
360 |
|
|
361 |
|
|
|
/* |
|
|
chroma optimize filter, invented by mf |
|
|
a chroma pixel is average from the surrounding pixels, when the |
|
|
correpsonding luma pixels are pure black or white. |
|
|
*/ |
|
|
|
|
|
void |
|
|
image_chroma_optimize(IMAGE * img, int width, int height, int edged_width) |
|
|
{ |
|
|
int x,y; |
|
|
int pixels = 0; |
|
|
|
|
|
for (y = 1; y < height/2 - 1; y++) |
|
|
for (x = 1; x < width/2 - 1; x++) |
|
|
{ |
|
|
#define IS_PURE(a) ((a)<=16||(a)>=235) |
|
|
#define IMG_Y(Y,X) img->y[(Y)*edged_width + (X)] |
|
|
#define IMG_U(Y,X) img->u[(Y)*edged_width/2 + (X)] |
|
|
#define IMG_V(Y,X) img->v[(Y)*edged_width/2 + (X)] |
|
|
|
|
|
if (IS_PURE(IMG_Y(y*2 ,x*2 )) && |
|
|
IS_PURE(IMG_Y(y*2 ,x*2+1)) && |
|
|
IS_PURE(IMG_Y(y*2+1,x*2 )) && |
|
|
IS_PURE(IMG_Y(y*2+1,x*2+1))) |
|
|
{ |
|
|
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; |
|
|
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; |
|
|
pixels++; |
|
|
} |
|
|
|
|
|
#undef IS_PURE |
|
|
#undef IMG_Y |
|
|
#undef IMG_U |
|
|
#undef IMG_V |
|
|
} |
|
|
|
|
|
DPRINTF(DPRINTF_DEBUG,"chroma_optimized_pixels = %i/%i", pixels, width*height/4); |
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* |
|
|
perform safe packed colorspace conversion, by splitting |
|
|
the image up into an optimized area (pixel width divisible by 16), |
|
|
and two unoptimized/plain-c areas (pixel width divisible by 2) |
|
|
*/ |
|
|
|
|
|
static void |
|
|
safe_packed_conv(uint8_t * x_ptr, int x_stride, |
|
|
uint8_t * y_ptr, uint8_t * u_ptr, uint8_t * v_ptr, |
|
|
int y_stride, int uv_stride, |
|
|
int width, int height, int vflip, |
|
|
packedFunc * func_opt, packedFunc func_c, int size) |
|
|
{ |
|
|
int width_opt, width_c; |
|
|
|
|
|
if (func_opt != func_c && x_stride < size*((width+15)/16)*16) |
|
|
{ |
|
|
width_opt = width & (~15); |
|
|
width_c = width - width_opt; |
|
|
} |
|
|
else |
|
|
{ |
|
|
width_opt = width; |
|
|
width_c = 0; |
|
|
} |
|
|
|
|
|
func_opt(x_ptr, x_stride, |
|
|
y_ptr, u_ptr, v_ptr, y_stride, uv_stride, |
|
|
width_opt, height, vflip); |
|
|
|
|
|
if (width_c) |
|
|
{ |
|
|
func_c(x_ptr + size*width_opt, x_stride, |
|
|
y_ptr + width_opt, u_ptr + width_opt/2, v_ptr + width_opt/2, |
|
|
y_stride, uv_stride, width_c, height, vflip); |
|
|
} |
|
|
} |
|
|
|
|
|
|
|
|
|
|
362 |
int |
int |
363 |
image_input(IMAGE * image, |
image_input(IMAGE * image, |
364 |
uint32_t width, |
uint32_t width, |
365 |
int height, |
int height, |
366 |
uint32_t edged_width, |
uint32_t edged_width, |
367 |
uint8_t * src, |
uint8_t * src, |
368 |
int src_stride, |
int csp) |
|
int csp, |
|
|
int interlacing) |
|
369 |
{ |
{ |
|
const int edged_width2 = edged_width/2; |
|
|
const int width2 = width/2; |
|
|
const int height2 = height/2; |
|
|
//const int height_signed = (csp & XVID_CSP_VFLIP) ? -height : height; |
|
370 |
|
|
371 |
|
/* if (csp & XVID_CSP_VFLIP) |
|
// int src_stride = width; |
|
|
|
|
|
// --- xvid 2.1 compatiblity patch --- |
|
|
// --- remove when xvid_dec_frame->stride equals real stride |
|
|
/* |
|
|
if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB555 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB565 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_YUY2 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_YVYU || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_UYVY) |
|
|
{ |
|
|
src_stride *= 2; |
|
|
} |
|
|
else if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB24) |
|
|
{ |
|
|
src_stride *= 3; |
|
|
} |
|
|
else if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB32 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_ABGR || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGBA) |
|
372 |
{ |
{ |
373 |
src_stride *= 4; |
height = -height; |
374 |
} |
} |
375 |
*/ |
*/ |
|
// ^--- xvid 2.1 compatiblity fix ---^ |
|
376 |
|
|
377 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
378 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
379 |
safe_packed_conv( |
rgb555_to_yv12(image->y, image->u, image->v, src, width, height, |
380 |
src, src_stride, image->y, image->u, image->v, |
edged_width); |
381 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
return 0; |
|
interlacing?rgb555i_to_yv12 :rgb555_to_yv12, |
|
|
interlacing?rgb555i_to_yv12_c:rgb555_to_yv12_c, 2); |
|
|
break; |
|
382 |
|
|
383 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
384 |
safe_packed_conv( |
rgb565_to_yv12(image->y, image->u, image->v, src, width, height, |
385 |
src, src_stride, image->y, image->u, image->v, |
edged_width); |
386 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
return 0; |
|
interlacing?rgb565i_to_yv12 :rgb565_to_yv12, |
|
|
interlacing?rgb565i_to_yv12_c:rgb565_to_yv12_c, 2); |
|
|
break; |
|
387 |
|
|
388 |
|
|
389 |
case XVID_CSP_RGB24: |
case XVID_CSP_RGB24: |
390 |
safe_packed_conv( |
rgb24_to_yv12(image->y, image->u, image->v, src, width, height, |
391 |
src, src_stride, image->y, image->u, image->v, |
edged_width); |
392 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
return 0; |
|
interlacing?bgri_to_yv12 :bgr_to_yv12, |
|
|
interlacing?bgri_to_yv12_c:bgr_to_yv12_c, 3); |
|
|
break; |
|
393 |
|
|
394 |
case XVID_CSP_RGB32: |
case XVID_CSP_RGB32: |
395 |
safe_packed_conv( |
rgb32_to_yv12(image->y, image->u, image->v, src, width, height, |
396 |
src, src_stride, image->y, image->u, image->v, |
edged_width); |
397 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
return 0; |
|
interlacing?bgrai_to_yv12 :bgra_to_yv12, |
|
|
interlacing?bgrai_to_yv12_c:bgra_to_yv12_c, 4); |
|
|
break; |
|
398 |
|
|
399 |
case XVID_CSP_ABGR : |
case XVID_CSP_I420: |
400 |
safe_packed_conv( |
yuv_to_yv12(image->y, image->u, image->v, src, width, height, |
401 |
src, src_stride, image->y, image->u, image->v, |
edged_width); |
402 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
return 0; |
|
interlacing?abgri_to_yv12 :abgr_to_yv12, |
|
|
interlacing?abgri_to_yv12_c:abgr_to_yv12_c, 4); |
|
|
break; |
|
403 |
|
|
404 |
case XVID_CSP_RGBA : |
case XVID_CSP_YV12: /* u/v swapped */ |
405 |
safe_packed_conv( |
yuv_to_yv12(image->y, image->v, image->u, src, width, height, |
406 |
src, src_stride, image->y, image->u, image->v, |
edged_width); |
407 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
return 0; |
|
interlacing?rgbai_to_yv12 :rgba_to_yv12, |
|
|
interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4); |
|
|
break; |
|
408 |
|
|
409 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
410 |
safe_packed_conv( |
yuyv_to_yv12(image->y, image->u, image->v, src, width, height, |
411 |
src, src_stride, image->y, image->u, image->v, |
edged_width); |
412 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
return 0; |
|
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
|
|
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
|
|
break; |
|
413 |
|
|
414 |
case XVID_CSP_YVYU: /* u/v swapped */ |
case XVID_CSP_YVYU: /* u/v swapped */ |
415 |
safe_packed_conv( |
yuyv_to_yv12(image->y, image->v, image->u, src, width, height, |
416 |
src, src_stride, image->y, image->v, image->y, |
edged_width); |
417 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
return 0; |
|
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
|
|
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
|
|
break; |
|
418 |
|
|
419 |
case XVID_CSP_UYVY: |
case XVID_CSP_UYVY: |
420 |
safe_packed_conv( |
uyvy_to_yv12(image->y, image->u, image->v, src, width, height, |
421 |
src, src_stride, image->y, image->u, image->v, |
edged_width); |
422 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
return 0; |
|
interlacing?uyvyi_to_yv12 :uyvy_to_yv12, |
|
|
interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2); |
|
|
break; |
|
|
|
|
|
case XVID_CSP_I420: |
|
|
yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
|
|
src, src + src_stride*height, src + src_stride*height + (src_stride/2)*height2, |
|
|
src_stride, src_stride/2, width, height, (csp & XVID_CSP_VFLIP)); |
|
|
break |
|
|
; |
|
|
case XVID_CSP_YV12: /* u/v swapped */ |
|
|
yv12_to_yv12(image->y, image->v, image->u, edged_width, edged_width2, |
|
|
src, src + src_stride*height, src + src_stride*height + (src_stride/2)*height2, |
|
|
src_stride, src_stride/2, width, height, (csp & XVID_CSP_VFLIP)); |
|
|
break; |
|
423 |
|
|
424 |
case XVID_CSP_USER: |
case XVID_CSP_USER: |
425 |
{ |
user_to_yuv_c(image->y, image->u, image->v, edged_width, |
426 |
DEC_PICTURE * pic = (DEC_PICTURE*)src; |
(DEC_PICTURE *) src, width, height); |
427 |
yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
return 0; |
|
pic->y, pic->u, pic->v, pic->stride_y, pic->stride_y, |
|
|
width, height, (csp & XVID_CSP_VFLIP)); |
|
|
} |
|
|
break; |
|
428 |
|
|
429 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
430 |
break; |
break; |
431 |
|
|
|
default : |
|
|
return -1; |
|
|
} |
|
|
|
|
|
|
|
|
/* pad out image when the width and/or height is not a multiple of 16 */ |
|
|
|
|
|
if (width & 15) |
|
|
{ |
|
|
int i; |
|
|
int pad_width = 16 - (width&15); |
|
|
for (i = 0; i < height; i++) |
|
|
{ |
|
|
memset(image->y + i*edged_width + width, |
|
|
*(image->y + i*edged_width + width - 1), pad_width); |
|
|
} |
|
|
for (i = 0; i < height/2; i++) |
|
|
{ |
|
|
memset(image->u + i*edged_width2 + width2, |
|
|
*(image->u + i*edged_width2 + width2 - 1),pad_width/2); |
|
|
memset(image->v + i*edged_width2 + width2, |
|
|
*(image->v + i*edged_width2 + width2 - 1),pad_width/2); |
|
|
} |
|
|
} |
|
|
|
|
|
if (height & 15) |
|
|
{ |
|
|
int pad_height = 16 - (height&15); |
|
|
int length = ((width+15)/16)*16; |
|
|
int i; |
|
|
for (i = 0; i < pad_height; i++) |
|
|
{ |
|
|
memcpy(image->y + (height+i)*edged_width, |
|
|
image->y + (height-1)*edged_width,length); |
|
|
} |
|
|
|
|
|
for (i = 0; i < pad_height/2; i++) |
|
|
{ |
|
|
memcpy(image->u + (height2+i)*edged_width2, |
|
|
image->u + (height2-1)*edged_width2,length/2); |
|
|
memcpy(image->v + (height2+i)*edged_width2, |
|
|
image->v + (height2-1)*edged_width2,length/2); |
|
|
} |
|
432 |
} |
} |
433 |
|
|
434 |
/* |
return -1; |
|
if (interlacing) |
|
|
image_printf(image, edged_width, height, 5,5, "[i]"); |
|
|
image_dump_yuvpgm(image, edged_width, ((width+15)/16)*16, ((height+15)/16)*16, "\\encode.pgm"); |
|
|
*/ |
|
|
return 0; |
|
435 |
} |
} |
436 |
|
|
437 |
|
|
443 |
uint32_t edged_width, |
uint32_t edged_width, |
444 |
uint8_t * dst, |
uint8_t * dst, |
445 |
uint32_t dst_stride, |
uint32_t dst_stride, |
446 |
int csp, |
int csp) |
|
int interlacing) |
|
|
{ |
|
|
const int edged_width2 = edged_width/2; |
|
|
int height2 = height/2; |
|
|
|
|
|
/* |
|
|
if (interlacing) |
|
|
image_printf(image, edged_width, height, 5,100, "[i]=%i,%i",width,height); |
|
|
image_dump_yuvpgm(image, edged_width, width, height, "\\decode.pgm"); |
|
|
*/ |
|
|
|
|
|
|
|
|
// --- xvid 2.1 compatiblity patch --- |
|
|
// --- remove when xvid_dec_frame->stride equals real stride |
|
|
/* |
|
|
if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB555 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB565 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_YUY2 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_YVYU || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_UYVY) |
|
|
{ |
|
|
dst_stride *= 2; |
|
|
} |
|
|
else if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB24) |
|
|
{ |
|
|
dst_stride *= 3; |
|
|
} |
|
|
else if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB32 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_ABGR || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGBA) |
|
447 |
{ |
{ |
448 |
dst_stride *= 4; |
if (csp & XVID_CSP_VFLIP) { |
449 |
|
height = -height; |
450 |
} |
} |
|
*/ |
|
|
// ^--- xvid 2.1 compatiblity fix ---^ |
|
|
|
|
451 |
|
|
452 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
453 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
454 |
safe_packed_conv( |
yv12_to_rgb555(dst, dst_stride, image->y, image->u, image->v, |
455 |
dst, dst_stride, image->y, image->u, image->v, |
edged_width, edged_width / 2, width, height); |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
|
|
interlacing?yv12_to_rgb555i :yv12_to_rgb555, |
|
|
interlacing?yv12_to_rgb555i_c:yv12_to_rgb555_c, 2); |
|
456 |
return 0; |
return 0; |
457 |
|
|
458 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
459 |
safe_packed_conv( |
yv12_to_rgb565(dst, dst_stride, image->y, image->u, image->v, |
460 |
dst, dst_stride, image->y, image->u, image->v, |
edged_width, edged_width / 2, width, height); |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
|
|
interlacing?yv12_to_rgb565i :yv12_to_rgb565, |
|
|
interlacing?yv12_to_rgb565i_c:yv12_to_rgb565_c, 2); |
|
461 |
return 0; |
return 0; |
462 |
|
|
463 |
case XVID_CSP_RGB24: |
case XVID_CSP_RGB24: |
464 |
safe_packed_conv( |
yv12_to_rgb24(dst, dst_stride, image->y, image->u, image->v, |
465 |
dst, dst_stride, image->y, image->u, image->v, |
edged_width, edged_width / 2, width, height); |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
|
|
interlacing?yv12_to_bgri :yv12_to_bgr, |
|
|
interlacing?yv12_to_bgri_c:yv12_to_bgr_c, 3); |
|
466 |
return 0; |
return 0; |
467 |
|
|
468 |
case XVID_CSP_RGB32: |
case XVID_CSP_RGB32: |
469 |
safe_packed_conv( |
yv12_to_rgb32(dst, dst_stride, image->y, image->u, image->v, |
470 |
dst, dst_stride, image->y, image->u, image->v, |
edged_width, edged_width / 2, width, height); |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
|
|
interlacing?yv12_to_bgrai :yv12_to_bgra, |
|
|
interlacing?yv12_to_bgrai_c:yv12_to_bgra_c, 4); |
|
|
return 0; |
|
|
|
|
|
case XVID_CSP_ABGR: |
|
|
safe_packed_conv( |
|
|
dst, dst_stride, image->y, image->u, image->v, |
|
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
|
|
interlacing?yv12_to_abgri :yv12_to_abgr, |
|
|
interlacing?yv12_to_abgri_c:yv12_to_abgr_c, 4); |
|
|
return 0; |
|
|
|
|
|
case XVID_CSP_RGBA: |
|
|
safe_packed_conv( |
|
|
dst, dst_stride, image->y, image->u, image->v, |
|
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
|
|
interlacing?yv12_to_rgbai :yv12_to_rgba, |
|
|
interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4); |
|
471 |
return 0; |
return 0; |
472 |
|
|
473 |
case XVID_CSP_YUY2: |
case XVID_CSP_I420: |
474 |
safe_packed_conv( |
yv12_to_yuv(dst, dst_stride, image->y, image->u, image->v, edged_width, |
475 |
dst, dst_stride, image->y, image->u, image->v, |
edged_width / 2, width, height); |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
|
|
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
|
|
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
|
476 |
return 0; |
return 0; |
477 |
|
|
478 |
case XVID_CSP_YVYU: // u,v swapped |
case XVID_CSP_YV12: // u,v swapped |
479 |
safe_packed_conv( |
yv12_to_yuv(dst, dst_stride, image->y, image->v, image->u, edged_width, |
480 |
dst, dst_stride, image->y, image->v, image->u, |
edged_width / 2, width, height); |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
|
|
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
|
|
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
|
481 |
return 0; |
return 0; |
482 |
|
|
483 |
case XVID_CSP_UYVY: |
case XVID_CSP_YUY2: |
484 |
safe_packed_conv( |
yv12_to_yuyv(dst, dst_stride, image->y, image->u, image->v, |
485 |
dst, dst_stride, image->y, image->u, image->v, |
edged_width, edged_width / 2, width, height); |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
|
|
interlacing?yv12_to_uyvyi :yv12_to_uyvy, |
|
|
interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2); |
|
486 |
return 0; |
return 0; |
487 |
|
|
488 |
case XVID_CSP_I420: |
case XVID_CSP_YVYU: // u,v swapped |
489 |
yv12_to_yv12(dst, dst + dst_stride*height, dst + dst_stride*height + (dst_stride/2)*height2, |
yv12_to_yuyv(dst, dst_stride, image->y, image->v, image->u, |
490 |
dst_stride, dst_stride/2, |
edged_width, edged_width / 2, width, height); |
|
image->y, image->u, image->v, edged_width, edged_width2, |
|
|
width, height, (csp & XVID_CSP_VFLIP)); |
|
491 |
return 0; |
return 0; |
492 |
|
|
493 |
case XVID_CSP_YV12: // u,v swapped |
case XVID_CSP_UYVY: |
494 |
yv12_to_yv12(dst, dst + dst_stride*height, dst + dst_stride*height + (dst_stride/2)*height2, |
yv12_to_uyvy(dst, dst_stride, image->y, image->u, image->v, |
495 |
dst_stride, dst_stride/2, |
edged_width, edged_width / 2, width, height); |
|
image->y, image->v, image->u, edged_width, edged_width2, |
|
|
width, height, (csp & XVID_CSP_VFLIP)); |
|
496 |
return 0; |
return 0; |
497 |
|
|
498 |
case XVID_CSP_USER: |
case XVID_CSP_USER: |
499 |
{ |
((DEC_PICTURE *) dst)->y = image->y; |
500 |
DEC_PICTURE * pic = (DEC_PICTURE*)dst; |
((DEC_PICTURE *) dst)->u = image->u; |
501 |
pic->y = image->y; |
((DEC_PICTURE *) dst)->v = image->v; |
502 |
pic->u = image->u; |
((DEC_PICTURE *) dst)->stride_y = edged_width; |
503 |
pic->v = image->v; |
((DEC_PICTURE *) dst)->stride_uv = edged_width / 2; |
|
pic->stride_y = edged_width; |
|
|
pic->stride_uv = edged_width / 2; |
|
|
} |
|
504 |
return 0; |
return 0; |
505 |
|
|
506 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
544 |
return psnr_y; |
return psnr_y; |
545 |
} |
} |
546 |
|
|
|
|
|
|
float sse_to_PSNR(long sse, int pixels) |
|
|
{ |
|
|
if (sse==0) |
|
|
return 99.99; |
|
|
|
|
|
return 48.131 - 10*log10((float)sse/(float)(pixels)); // log10(255*255)=4.8131 |
|
|
|
|
|
} |
|
|
|
|
|
long plane_sse(uint8_t * orig, |
|
|
uint8_t * recon, |
|
|
uint16_t stride, |
|
|
uint16_t width, |
|
|
uint16_t height) |
|
|
{ |
|
|
int diff, x, y; |
|
|
long sse=0; |
|
|
|
|
|
for (y = 0; y < height; y++) { |
|
|
for (x = 0; x < width; x++) { |
|
|
diff = *(orig + x) - *(recon + x); |
|
|
sse += diff * diff; |
|
|
} |
|
|
orig += stride; |
|
|
recon += stride; |
|
|
} |
|
|
return sse; |
|
|
} |
|
|
|
|
547 |
/* |
/* |
548 |
|
|
549 |
#include <stdio.h> |
#include <stdio.h> |
640 |
} |
} |
641 |
|
|
642 |
|
|
643 |
|
#define ABS(X) (((X)>0)?(X):-(X)) |
644 |
float |
float |
645 |
image_mad(const IMAGE * img1, |
image_mad(const IMAGE * img1, |
646 |
const IMAGE * img2, |
const IMAGE * img2, |
703 |
sV += std2; |
sV += std2; |
704 |
} |
} |
705 |
} |
} |
|
|
|
|
|
|
|
void |
|
|
image_clear(IMAGE * img, int width, int height, int edged_width, |
|
|
int y, int u, int v) |
|
|
{ |
|
|
uint8_t * p; |
|
|
int i; |
|
|
|
|
|
p = img->y; |
|
|
for (i = 0; i < height; i++) { |
|
|
memset(p, y, width); |
|
|
p += edged_width; |
|
|
} |
|
|
|
|
|
p = img->u; |
|
|
for (i = 0; i < height/2; i++) { |
|
|
memset(p, u, width/2); |
|
|
p += edged_width/2; |
|
|
} |
|
|
|
|
|
p = img->v; |
|
|
for (i = 0; i < height/2; i++) { |
|
|
memset(p, v, width/2); |
|
|
p += edged_width/2; |
|
|
} |
|
|
} |
|
|
|
|
|
|
|
|
/* reduced resolution deblocking filter |
|
|
block = block size (16=rrv, 8=full resolution) |
|
|
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; |
|
|
|
|
|
/* luma: j,i in block units */ |
|
|
if ((flags & XVID_DEC_DEBLOCKY)) |
|
|
{ |
|
|
for (j = 1; j < mb_height*2; j++) /* horizontal deblocking */ |
|
|
for (i = 0; i < mb_width*2; i++) |
|
|
{ |
|
|
if (mbs[(j-1)/2*mb_stride + (i/2)].mode != MODE_NOT_CODED || |
|
|
mbs[(j+0)/2*mb_stride + (i/2)].mode != MODE_NOT_CODED) |
|
|
{ |
|
|
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++) |
|
|
{ |
|
|
if (mbs[(j/2)*mb_stride + (i-1)/2].mode != MODE_NOT_CODED || |
|
|
mbs[(j/2)*mb_stride + (i+0)/2].mode != MODE_NOT_CODED) |
|
|
{ |
|
|
vfilter_31(img->y + (j*block)*edged_width + i*block - 1, |
|
|
img->y + (j*block)*edged_width + i*block + 0, |
|
|
edged_width, nblocks); |
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
|
|
|
/* chroma */ |
|
|
if ((flags & XVID_DEC_DEBLOCKUV)) |
|
|
{ |
|
|
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); |
|
|
} |
|
|
} |
|
|
|
|
|
for (j = 0; j < mb_height; j++) /* vertical deblocking */ |
|
|
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) |
|
|
{ |
|
|
vfilter_31(img->u + (j*block)*edged_width2 + i*block - 1, |
|
|
img->u + (j*block)*edged_width2 + i*block + 0, |
|
|
edged_width2, nblocks); |
|
|
vfilter_31(img->v + (j*block)*edged_width2 + i*block - 1, |
|
|
img->v + (j*block)*edged_width2 + i*block + 0, |
|
|
edged_width2, nblocks); |
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
} |
|
|
|
|