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" |
#include "../motion/sad.h" |
37 |
|
|
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; |
252 |
#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 |
|
253 |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
254 |
uint32_t x, y; |
uint32_t x, y; |
255 |
|
|
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 |
|
|
595 |
int height, |
int height, |
596 |
uint32_t edged_width, |
uint32_t edged_width, |
597 |
uint8_t * dst[4], |
uint8_t * dst[4], |
598 |
uint32_t dst_stride[4], |
int dst_stride[4], |
599 |
int csp, |
int csp, |
600 |
int interlacing) |
int interlacing) |
601 |
{ |
{ |
956 |
} |
} |
957 |
|
|
958 |
void |
void |
959 |
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) { |
960 |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
961 |
int std2 = std >> 1; |
int stride2 = stride >> 1; |
962 |
int w = mbl << 4, w2,i; |
int w = mbl << 4, w2,i; |
963 |
|
|
964 |
if(w > width) |
if(w > width) |
968 |
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); |
969 |
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); |
970 |
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); |
971 |
sY = cur->y + (mby << 4) * std + (mbx << 4); |
sY = cur->y + (mby << 4) * stride + (mbx << 4); |
972 |
sU = cur->u + (mby << 3) * std2 + (mbx << 3); |
sU = cur->u + (mby << 3) * stride2 + (mbx << 3); |
973 |
sV = cur->v + (mby << 3) * std2 + (mbx << 3); |
sV = cur->v + (mby << 3) * stride2 + (mbx << 3); |
974 |
|
|
975 |
for(i = 0 ; i < 16 ; i++) { |
for(i = 0 ; i < 16 ; i++) { |
976 |
memcpy(dY,sY,w); |
memcpy(dY,sY,w); |
977 |
dY += out_frm->stride[0]; |
dY += out_frm->stride[0]; |
978 |
sY += std; |
sY += stride; |
979 |
} |
} |
980 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
981 |
memcpy(dU,sU,w2); |
memcpy(dU,sU,w2); |
982 |
dU += out_frm->stride[1]; |
dU += out_frm->stride[1]; |
983 |
sU += std2; |
sU += stride2; |
984 |
} |
} |
985 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
986 |
memcpy(dV,sV,w2); |
memcpy(dV,sV,w2); |
987 |
dV += out_frm->stride[2]; |
dV += out_frm->stride[2]; |
988 |
sV += std2; |
sV += stride2; |
989 |
} |
} |
990 |
} |
} |
991 |
|
|
1015 |
p += edged_width/2; |
p += edged_width/2; |
1016 |
} |
} |
1017 |
} |
} |
|
|
|
|
|
|
|
/* 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 */ |
|
|
|
|
|
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 */ |
|
|
|
|
|
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); |
|
|
} |
|
|
} |
|
|
|
|
|
|
|
|
} |
|
|
|
|