提交 1f26c6f3 编写于 作者: M Michael Niedermayer

rate distored optimal lambda->qp support

Originally committed as revision 2509 to svn://svn.ffmpeg.org/ffmpeg/trunk
上级 567e36d8
......@@ -255,6 +255,7 @@ static const __attribute__((unused)) int Motion_Est_QTab[] =
#define CODEC_FLAG_AC_PRED 0x01000000 ///< H263 Advanced intra coding / MPEG4 AC prediction
#define CODEC_FLAG_H263P_UMV 0x02000000 ///< Unlimited motion vector
#define CODEC_FLAG_CBP_RD 0x04000000 ///< use rate distortion optimization for cbp
#define CODEC_FLAG_QP_RD 0x08000000 ///< use rate distortion optimization for qp selectioon
/* For advanced prediction mode, we reuse the 4MV flag */
/* Unsupported options :
* Syntax Arithmetic coding (SAC)
......@@ -1065,6 +1066,7 @@ typedef struct AVCodecContext {
/**
* sample aspect ratio (0 if unknown).
* numerator and denominator must be relative prime and smaller then 256 for some video standards
* - encoding: set by user.
* - decoding: set by lavc.
*/
......
......@@ -664,7 +664,8 @@ int MPV_encode_init(AVCodecContext *avctx)
|| s->avctx->dark_masking
|| s->avctx->temporal_cplx_masking
|| s->avctx->spatial_cplx_masking
|| s->avctx->p_masking)
|| s->avctx->p_masking
|| (s->flags&CODEC_FLAG_QP_RD))
&& !s->fixed_qscale;
s->progressive_sequence= !(avctx->flags & CODEC_FLAG_INTERLACED_DCT);
......@@ -2898,15 +2899,18 @@ static void encode_mb(MpegEncContext *s, int motion_x, int motion_y)
s->lambda= s->lambda_table[mb_xy];
update_qscale(s);
s->dquant= s->qscale - last_qp;
if(!(s->flags&CODEC_FLAG_QP_RD)){
s->dquant= s->qscale - last_qp;
if(s->out_format==FMT_H263)
s->dquant= clip(s->dquant, -2, 2); //FIXME RD
if(s->out_format==FMT_H263)
s->dquant= clip(s->dquant, -2, 2); //FIXME RD
if(s->codec_id==CODEC_ID_MPEG4){
if(!s->mb_intra){
if((s->mv_dir&MV_DIRECT) || s->mv_type==MV_TYPE_8X8)
s->dquant=0;
if(s->codec_id==CODEC_ID_MPEG4){
if(!s->mb_intra){
if((s->mv_dir&MV_DIRECT) || s->mv_type==MV_TYPE_8X8)
s->dquant=0;
}
}
}
s->qscale= last_qp + s->dquant;
......@@ -3256,6 +3260,7 @@ static inline void copy_context_before_encode(MpegEncContext *d, MpegEncContext
d->mb_skiped= 0;
d->qscale= s->qscale;
d->dquant= s->dquant;
}
static inline void copy_context_after_encode(MpegEncContext *d, MpegEncContext *s, int type){
......@@ -3724,8 +3729,9 @@ static void encode_picture(MpegEncContext *s, int picture_number)
}
s->mb_skiped=0;
s->dquant=0; //only for QP_RD
if(mb_type & (mb_type-1)){ // more than 1 MB type possible
if(mb_type & (mb_type-1) || (s->flags & CODEC_FLAG_QP_RD)){ // more than 1 MB type possible
int next_block=0;
int pb_bits_count, pb2_bits_count, tex_pb_bits_count;
......@@ -3823,6 +3829,52 @@ static void encode_picture(MpegEncContext *s, int picture_number)
ff_clean_intra_table_entries(s); //old mode?
}
}
if(s->flags & CODEC_FLAG_QP_RD){
if(best_s.mv_type==MV_TYPE_16X16 && !(best_s.mv_dir&MV_DIRECT)){
const int last_qp= backup_s.qscale;
int dquant, dir, qp, dc[6];
assert(backup_s.dquant == 0);
//FIXME intra
s->mv_dir= best_s.mv_dir;
s->mv_type = MV_TYPE_16X16;
s->mb_intra= best_s.mb_intra;
s->mv[0][0][0] = best_s.mv[0][0][0];
s->mv[0][0][1] = best_s.mv[0][0][1];
s->mv[1][0][0] = best_s.mv[1][0][0];
s->mv[1][0][1] = best_s.mv[1][0][1];
dir= s->pict_type == B_TYPE ? 2 : 1;
if(last_qp + dir >= s->avctx->qmax) dir= -dir;
for(dquant= dir; dquant<=2 && dquant>=-2; dquant += dir){
qp= last_qp + dquant;
if(qp < s->avctx->qmin || qp > s->avctx->qmax)
break;
backup_s.dquant= dquant;
for(i=0; i<6; i++){
dc[i]= s->dc_val[0][ s->block_index[i] ]; //FIXME AC
}
//printf("%d %d\n", backup_s.dquant, backup_s.qscale);
encode_mb_hq(s, &backup_s, &best_s, MB_TYPE_INTER /* wrong but unused */, pb, pb2, tex_pb,
&dmin, &next_block, s->mv[0][0][0], s->mv[0][0][1]);
if(best_s.qscale != qp){
for(i=0; i<6; i++){
s->dc_val[0][ s->block_index[i] ]= dc[i];
}
if(dir > 0 && dquant==dir){
dquant= 0;
dir= -dir;
}else
break;
}
}
qp= best_s.qscale;
s->current_picture.qscale_table[xy]= qp;
}
}
copy_context_after_encode(s, &best_s, -1);
pb_bits_count= get_bit_count(&s->pb);
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册