当前位置: 首页 > 工具软件 > x264-go > 使用案例 >

帧类型决策-x264_slicetype_analyse()

微生博简
2023-12-01
/*
	对lookahead队列中的帧分析,确定他们的帧类型

	过程:
		1.	若第一帧是AUTO/I,检查其相对于上一个非B帧是否场景切换,若场景切换则将当前帧设置为I,return
		2.	遍历,将所有的关键帧根据openGOP设置成I/IDR
		3.	将所有IDR帧前的AUTO/B帧改为P帧
		4.	若允许使用B帧
			·	使用X264_B_ADAPT_TRELLIS方法自适应设置B帧
			·	使用X264_B_ADAPT_FAST方法自适应设置B帧
			·	不自适应设置B帧
			若不允许使用B帧,则将全部的AUTO/B设置成P
		5.	将最后一帧AUTO/B帧改为P帧
		6.	若允许宏块树分析则进行宏块树分析
		7.	根据关键帧之间的距离限制检查帧类型
		8.	若允许vbv则进行vbv检查
*/
void x264_slicetype_analyse( x264_t *h, int intra_minigop )
{
    x264_mb_analysis_t a;
    x264_frame_t *frames[X264_LOOKAHEAD_MAX+3] = { NULL, };
    int num_frames, orig_num_frames, keyint_limit, framecnt;
    int i_max_search = X264_MIN( h->lookahead->next.i_size, X264_LOOKAHEAD_MAX );
    int b_vbv_lookahead = h->param.rc.i_vbv_buffer_size && h->param.rc.i_lookahead;
    /* For determinism we should limit the search to the number of frames lookahead has for sure
     * in h->lookahead->next.list buffer, except at the end of stream.
     * For normal calls with (intra_minigop == 0) that is h->lookahead->i_slicetype_length + 1 frames.
     * And for I-frame calls (intra_minigop != 0) we already removed intra_minigop frames from there. */
    if( h->param.b_deterministic )
        i_max_search = X264_MIN( i_max_search, h->lookahead->i_slicetype_length + 1 - intra_minigop );
    int keyframe = !!intra_minigop;

    assert( h->frames.b_have_lowres );

    if( !h->lookahead->last_nonb )	//若前面没有非B帧,则return
        return;

	//frame[0]存储前面最近的非B帧,frame[1~i_max_search]存储要分析的帧
    frames[0] = h->lookahead->last_nonb;	//取最近的非B帧为frame0
    for( framecnt = 0; framecnt < i_max_search; framecnt++ )	//从list中拷贝到frames中
        frames[framecnt+1] = h->lookahead->next.list[framecnt];

	//低分辨率上下文初始化
    lowres_context_init( h, &a );

    if( !framecnt )	//若framecnt = 0,即i_max_search = 0,则无帧分析,return
    {
        if( h->param.rc.b_mb_tree )
            macroblock_tree( h, &a, frames, 0, keyframe );
        return;
    }

	//计算keyint_limit = 最大关键帧距离 - 最近的非B帧 + 最近一个关键帧 - 1
    keyint_limit = h->param.i_keyint_max - frames[0]->i_frame + h->lookahead->i_last_keyframe - 1;

	//计算orig_num_frames和num_frames
    orig_num_frames = num_frames = h->param.b_intra_refresh ? framecnt : X264_MIN( framecnt, keyint_limit );

    /* This is important psy-wise: if we have a non-scenecut keyframe,
     * there will be significant visual artifacts if the frames just before
     * go down in quality due to being referenced less, despite it being
     * more RD-optimal. 
	 * 如果我们有一个非场景切换的关键帧,将会有一个明显的视觉影响
	 */
    if( (h->param.analyse.b_psy && h->param.rc.b_mb_tree) || b_vbv_lookahead )
        num_frames = framecnt;
    else if( h->param.b_open_gop && num_frames < framecnt )
        num_frames++;
    else if( num_frames == 0 )
    {
        frames[1]->i_type = X264_TYPE_I;
        return;
    }

    if( IS_X264_TYPE_AUTO_OR_I( frames[1]->i_type ) &&
        h->param.i_scenecut_threshold && scenecut( h, &a, frames, 0, 1, 1, orig_num_frames, i_max_search ) )
    {	//若是第一帧是auto/I 且 相对于上一个非B帧有场景切换,则设置为I帧
        if( frames[1]->i_type == X264_TYPE_AUTO )	
            frames[1]->i_type = X264_TYPE_I;	//当前帧定为I帧
        return;
    }

    /* Replace forced keyframes with I/IDR-frames */
    for( int j = 1; j <= num_frames; j++ )
    {	遍历检查是否关键帧,依openGOP设定为I/IDR
        if( frames[j]->i_type == X264_TYPE_KEYFRAME )
            frames[j]->i_type = h->param.b_open_gop ? X264_TYPE_I : X264_TYPE_IDR;
    }

    /* Close GOP at IDR-frames */
    for( int j = 2; j <= num_frames; j++ )
    {	//遍历检查IDR帧,将IDR帧前的AUTO/B帧设置为P帧
        if( frames[j]->i_type == X264_TYPE_IDR && IS_X264_TYPE_AUTO_OR_B( frames[j-1]->i_type ) )
            frames[j-1]->i_type = X264_TYPE_P;
    }

    int num_analysed_frames = num_frames;
    int reset_start;

    if( h->param.i_bframe )	//若允许B帧
    {
		/*	根据i_bframe_adaptive来进行自适应B帧设置
			i_bframe_adaptive = X264_B_ADAPT_TRELLIS	使用最佳判别方法,即viterbi
			i_bframe_adaptive = X264_B_ADAPT_FAST		使用快速判别方法
			i_bframe_adaptive = X264_B_ADAPT_NONE		不使用自适应B帧	*/
        if( h->param.i_bframe_adaptive == X264_B_ADAPT_TRELLIS )//b_adaptive=2,使用最佳算法,即viterbi
        {
            if( num_frames > 1 )
            {
				//初始化最优帧类型路径 和 路径索引
                char best_paths[X264_BFRAME_MAX+1][X264_LOOKAHEAD_MAX+1] = {"","P"};
                int best_path_index = num_frames % (X264_BFRAME_MAX+1);

                /* Perform the frametype analysis. */
                for( int j = 2; j <= num_frames; j++ )
					//执行viterbi计算最优帧类型路径,将最优帧类型路径写入best_paths
                    slicetype_path( h, &a, frames, j, best_paths );

                /* Load the results of the analysis into the frame types. */
				//根据分析的最优帧类型路径设置各个帧
                for( int j = 1; j < num_frames; j++ )
                {
                    if( best_paths[best_path_index][j-1] != 'B' )
                    {	//若是'P'则将AUTO/B该成P
                        if( IS_X264_TYPE_AUTO_OR_B( frames[j]->i_type ) )
                            frames[j]->i_type = X264_TYPE_P;
                    }
                    else
                    {	//若是'B'则将AUTO改成B
                        if( frames[j]->i_type == X264_TYPE_AUTO )
                            frames[j]->i_type = X264_TYPE_B;
                    }
                }
            }
        }	//end of b_adaptive = X264_B_ADAPT_TRELLIS

        else if( h->param.i_bframe_adaptive == X264_B_ADAPT_FAST )	//b_adaptive=1,快速决定
        {
            int last_nonb = 0;
            int num_bframes = h->param.i_bframe;	//得到两个参考帧中最多所允许的B帧数量
            char path[X264_LOOKAHEAD_MAX+1];
            for( int j = 1; j < num_frames; j++ )	//遍历list里的每一帧,除了最后一帧
            {
				/*
					针对前一帧来更新 B坑数num_bframes 和 最近非B帧last_nonb
				*/
                if( j-1 > 0 && IS_X264_TYPE_B( frames[j-1]->i_type ) )	//若前一帧为B帧
                    num_bframes--;	//则占一个B帧的坑
                else	//若前一帧为非B帧
                {
                    last_nonb = j-1;	//更新最近非B帧为前一帧
                    num_bframes = h->param.i_bframe;	//这重置B帧的坑数了
                }

				/*
					设置当前帧的帧类型
				*/
                if( !num_bframes )	//若B帧坑占满了,且当前帧B/auto则改P,其余不变
                {
                    if( IS_X264_TYPE_AUTO_OR_B( frames[j]->i_type ) )
                        frames[j]->i_type = X264_TYPE_P;
                    continue;
                }
				// else	允许到这里的帧允许使用B帧 (还有B坑)
                if( frames[j]->i_type != X264_TYPE_AUTO )	//若已经有类型了,则下一个
                    continue;
				// else	运行到这里的帧全是AUTO帧
                if( IS_X264_TYPE_B( frames[j+1]->i_type ) )	//若后面一帧为B帧
                {
                    frames[j]->i_type = X264_TYPE_P;	//当前帧改P,为什么??
                    continue;
                }
				/* else	运行到这里的帧后面一帧一定是P帧,不可能是I帧,
				   因为之前将所有I帧的前一帧设置成P帧,即只可能PI,不可能BI;  */

                int bframes = j - last_nonb - 1;	//得到上一个非B帧到最后一帧之间的B帧数量bframes
                memset( path, 'B', bframes );		//在path中填充bframes个B

				//	将当前帧设置为P帧,计算BBB...BBB(P)P帧类型路径的开销
                strcpy( path+bframes, "PP" );
                uint64_t cost_p = slicetype_path_cost( h, &a, frames+last_nonb, path, COST_MAX64 );

				//	将当前帧设置为B帧,计算BBB...BBB(B)P帧类型路径的开销
                strcpy( path+bframes, "BP" );		
                uint64_t cost_b = slicetype_path_cost( h, &a, frames+last_nonb, path, cost_p );

				//选择开销最小的,定当前帧的帧类型
                if( cost_b < cost_p )
                    frames[j]->i_type = X264_TYPE_B;
                else
                    frames[j]->i_type = X264_TYPE_P;
            }
        }	//end of b_adaptive = X264_B_ADAPT_FAST

        else	//b_adaptive=0,即不适用自适应B帧
        {
            int num_bframes = h->param.i_bframe;
            for( int j = 1; j < num_frames; j++ )
            {
                if( !num_bframes )	//没有B帧坑位了
                {
                    if( IS_X264_TYPE_AUTO_OR_B( frames[j]->i_type ) )
                        frames[j]->i_type = X264_TYPE_P;	//设置为P
                }
                else if( frames[j]->i_type == X264_TYPE_AUTO )	//有B帧坑位
                {
                    if( IS_X264_TYPE_B( frames[j+1]->i_type ) )	//后面是B则当前设P,同样为什么?
                        frames[j]->i_type = X264_TYPE_P;
                    else										//后面是P则当前设B
                        frames[j]->i_type = X264_TYPE_B;
                }

				//更新坑位
                if( IS_X264_TYPE_B( frames[j]->i_type ) )	//B帧
                    num_bframes--;							//占一个坑位
                else
                    num_bframes = h->param.i_bframe;		//更新坑位
            }
        }	//end of b_adaptive = X264_B_ADAPT_NONE		自适应B帧判别结束

        if( IS_X264_TYPE_AUTO_OR_B( frames[num_frames]->i_type ) )	//如果最后一帧是B,则改P
            frames[num_frames]->i_type = X264_TYPE_P;

		//从第1帧开始统计连续的B帧数量
        int num_bframes = 0;
        while( num_bframes < num_frames && IS_X264_TYPE_B( frames[num_bframes+1]->i_type ) )
            num_bframes++;

        /* Check scenecut on the first minigop. */
        for( int j = 1; j < num_bframes+1; j++ )	//遍历连续的B帧
        {
            if( frames[j]->i_forced_type == X264_TYPE_AUTO && IS_X264_TYPE_AUTO_OR_I( frames[j+1]->i_forced_type ) &&
                h->param.i_scenecut_threshold && scenecut( h, &a, frames, j, j+1, 0, orig_num_frames, i_max_search ) )
            {	//若后一帧是相对当前帧是场景切换,则当前帧设置为P帧
                frames[j]->i_type = X264_TYPE_P;
                num_analysed_frames = j;
                break;
            }
        }

        reset_start = keyframe ? 1 : X264_MIN( num_bframes+2, num_analysed_frames+1 );
    }
    else		//不允许B帧
    {
        for( int j = 1; j <= num_frames; j++ )
            if( IS_X264_TYPE_AUTO_OR_B( frames[j]->i_type ) )	//所有AUTO/B帧全改P
                frames[j]->i_type = X264_TYPE_P;
        reset_start = !keyframe + 1;
    }

    /* Perform the actual macroblock tree analysis.
     * Don't go farther than the maximum keyframe interval; this helps in short GOPs. */
    if( h->param.rc.b_mb_tree )	//使用mb tree
        macroblock_tree( h, &a, frames, X264_MIN(num_frames, h->param.i_keyint_max), keyframe );

    /* Enforce keyframe limit.	进行关键帧限制 */
    if( !h->param.b_intra_refresh )	//是否用定期的帧内刷新来替代IDR
    {
        int last_keyframe = h->lookahead->i_last_keyframe;	//取最近一个关键帧
        int last_possible = 0;
        for( int j = 1; j <= num_frames; j++ )	//遍历所有帧
        {
            x264_frame_t *frm = frames[j];
            int keyframe_dist = frm->i_frame - last_keyframe;	//计算距离上一个关键帧的距离

            if( IS_X264_TYPE_AUTO_OR_I( frm->i_forced_type ) )	//若i_forced_type是I或auto
            {
                if( h->param.b_open_gop || !IS_X264_TYPE_B( frames[j-1]->i_forced_type ) )//openGOP || 前一帧非B
                    last_possible = j;
            }
            if( keyframe_dist >= h->param.i_keyint_max )
            {	//若当前帧距离上一关键帧的距离 >= 最大关键帧间距离,则依据openGOP设置当前帧为I/IDR
                if( last_possible != 0 && last_possible != j )
                {
                    j = last_possible;
                    frm = frames[j];
                    keyframe_dist = frm->i_frame - last_keyframe;
                }
                last_possible = 0;
                if( frm->i_type != X264_TYPE_IDR )	//非IDR,根据openGOP设置I/IDR
                    frm->i_type = h->param.b_open_gop ? X264_TYPE_I : X264_TYPE_IDR;
            }
            if( frm->i_type == X264_TYPE_I && keyframe_dist >= h->param.i_keyint_min )
            {	//若是I帧 && 当前帧距离上一关键帧的距离 >= 最小关键帧间距离
                if( h->param.b_open_gop )	//openGOP
                {
                    last_keyframe = frm->i_frame;	//更新当前帧为上一关键帧
                    if( h->param.b_bluray_compat )
                    {
                        // Use bluray order
                        int bframes = 0;
                        while( bframes < j-1 && IS_X264_TYPE_B( frames[j-1-bframes]->i_type ) )
                            bframes++;
                        last_keyframe -= bframes;
                    }
                }
                else if( frm->i_forced_type != X264_TYPE_I )//非openGOP,且i_forced_type不是I帧,则设置为IDR
                    frm->i_type = X264_TYPE_IDR;
            }
            if( frm->i_type == X264_TYPE_IDR )	//若是IDR
            {
                last_keyframe = frm->i_frame;	//更新 最近关键帧
                if( j > 1 && IS_X264_TYPE_B( frames[j-1]->i_type ) )	//若前一帧是B
                    frames[j-1]->i_type = X264_TYPE_P;	//改为P
            }
        }
    }

    if( b_vbv_lookahead )	//若允许vbv,则进行vbv
        vbv_lookahead( h, &a, frames, num_frames, keyframe );

    /* Restore frametypes for all frames that haven't actually been decided yet. */
	//将需要重新设置帧类型的帧全部重置
    for( int j = reset_start; j <= num_frames; j++ )
        frames[j]->i_type = frames[j]->i_forced_type;
}
其中使用自适应B帧 X264_B_ADAPT_TRELLIS方法部分:
/* Viterbi/trellis slicetype decision algorithm. */
/* Uses strings due to the fact that the speed of the control functions is
   negligible compared to the cost of running slicetype_frame_cost, and because
   it makes debugging easier. 
   
   使用viterbi算法来确定frames[1~length]的最优帧类型路径
   将结果存放在best_paths[length % (X264_BFRAME_MAX+1)]中,以字符I/P/B表示
   */
static void slicetype_path( x264_t *h, x264_mb_analysis_t *a, x264_frame_t **frames, int length, char (*best_paths)[X264_LOOKAHEAD_MAX+1] )
{
    char paths[2][X264_LOOKAHEAD_MAX+1];

	//分析的路径数量最多i_bframe+1个,即最多向前分析到BBB...BBBP,length表示需要1~length帧,取两者最小值
    int num_paths = X264_MIN( h->param.i_bframe+1, length );
    
	uint64_t best_cost = COST_MAX64;	//初始化best_cost为MAX

    int best_possible = 0;
    int idx = 0;

    /* Iterate over all currently possible paths */
    for( int path = 0; path < num_paths; path++ )	//遍历每条路径
    {
        /* Add suffixes to the current path 构造路径 
		 * 结构:
		 *			  |			   length		   |
		 *		即    (	  best-path   ,B  ... B, P )
		 *		      |     len个     | path个 |1个|
		 * path range:
		 *			  |   len个                |1个|	path = 0
		 *			  |   len个            |1个|1个|		path = 1
		 *			  |   len个           | 2个|1个|		path = 2
		 *			  |   len个          | 3个 |1个|		path = 3
		 *							...						...
		 *			  |   len个  | num_paths个 |1个|		path = num_paths */
        int len = length - (path + 1);
        memcpy( paths[idx], best_paths[len % (X264_BFRAME_MAX+1)], len );//拷贝之前已经计算好的最优路径
        memset( paths[idx]+len, 'B', path );	//再到后面添加path个'B'
        strcpy( paths[idx]+len+path, "P" );		//最后一帧改成P帧

        int possible = 1;
        for( int i = 1; i <= length; i++ )	//遍历待确定类型的每一帧,来修改path
        {
            int i_type = frames[i]->i_type;
            if( i_type == X264_TYPE_AUTO )	//若auto则不改path
                continue;
            if( IS_X264_TYPE_B( i_type ) )	//若是B帧
				//该帧在已确定的最优路径内 || 该帧为待分析的最后一帧 || 该帧在路径中标识为B
                possible = possible && (i < len || i == length || paths[idx][i-1] == 'B');
            else	//若非B则改path
            {	
				//该帧在已确定的最优路径内 || 该帧在路径中标识为非B
                possible = possible && (i < len || paths[idx][i-1] != 'B');
                paths[idx][i-1] = IS_X264_TYPE_I( i_type ) ? 'I' : 'P';	//根据帧类型修改路径中的标识
            }
        }

        if( possible || !best_possible )	//possible和best_possible什么意思???
        {
            if( possible && !best_possible ) 
                best_cost = COST_MAX64;

            /* Calculate the actual cost of the current path 计算当前路径paths[idx]的开销*/
            uint64_t cost = slicetype_path_cost( h, a, frames, paths[idx], best_cost );

            if( cost < best_cost )	//更新best,且切换paths idx
            {
                best_cost = cost;
                best_possible = possible;
                idx ^= 1;
            }
        }
    }	//end of for( int path = 0; path < num_paths; path++ )

    /* Store the best path. 存储最优路径*/
    memcpy( best_paths[length % (X264_BFRAME_MAX+1)], paths[idx^1], length );
}
其中使用自适应B帧 X264_B_ADAPT_FAST方法部分:
/*
	过程:
		1.	根据前一帧的帧类型来更新B帧的坑位以及最近的非B帧
		2.	若B帧没坑位,则将AUTO/B改P,则continue
		3.	若已经确定了帧类型,即非AUTO,则continue
		4.	若后面一帧是B帧,设置当前帧为P帧
			若后面一帧是P帧,计算当前帧为B帧时和P帧时的帧类型路径开销,选开销小的作为当前帧类型
*/
if( h->param.i_bframe_adaptive == X264_B_ADAPT_FAST )	//b_adaptive=1,快速决定
        {
            int last_nonb = 0;
            int num_bframes = h->param.i_bframe;	//得到两个参考帧中最多所允许的B帧数量
            char path[X264_LOOKAHEAD_MAX+1];
            for( int j = 1; j < num_frames; j++ )	//遍历list里的每一帧,除了最后一帧
            {
				/*
					针对前一帧来更新 B坑数num_bframes 和 最近非B帧last_nonb
				*/
                if( j-1 > 0 && IS_X264_TYPE_B( frames[j-1]->i_type ) )	//若前一帧为B帧
                    num_bframes--;	//则占一个B帧的坑
                else	//若前一帧为非B帧
                {
                    last_nonb = j-1;	//更新最近非B帧为前一帧
                    num_bframes = h->param.i_bframe;	//这重置B帧的坑数了
                }

				/*
					设置当前帧的帧类型
				*/
                if( !num_bframes )	//若B帧坑占满了,且当前帧B/auto则改P,其余不变
                {
                    if( IS_X264_TYPE_AUTO_OR_B( frames[j]->i_type ) )
                        frames[j]->i_type = X264_TYPE_P;
                    continue;
                }
				// else	允许到这里的帧允许使用B帧 (还有B坑)
                if( frames[j]->i_type != X264_TYPE_AUTO )	//若已经有类型了,则下一个
                    continue;
				// else	运行到这里的帧全是AUTO帧
                if( IS_X264_TYPE_B( frames[j+1]->i_type ) )	//若后面一帧为B帧
                {
                    frames[j]->i_type = X264_TYPE_P;	//当前帧改P,为什么??
                    continue;
                }
				/* else	运行到这里的帧后面一帧一定是P帧,不可能是I帧,
				   因为之前将所有I帧的前一帧设置成P帧,即只可能PI,不可能BI;  */

                int bframes = j - last_nonb - 1;	//得到上一个非B帧到最后一帧之间的B帧数量bframes
                memset( path, 'B', bframes );		//在path中填充bframes个B

				//	将当前帧设置为P帧,计算BBB...BBB(P)P帧类型路径的开销
                strcpy( path+bframes, "PP" );
                uint64_t cost_p = slicetype_path_cost( h, &a, frames+last_nonb, path, COST_MAX64 );

				//	将当前帧设置为B帧,计算BBB...BBB(B)P帧类型路径的开销
                strcpy( path+bframes, "BP" );		
                uint64_t cost_b = slicetype_path_cost( h, &a, frames+last_nonb, path, cost_p );

				//选择开销最小的,定当前帧的帧类型
                if( cost_b < cost_p )
                    frames[j]->i_type = X264_TYPE_B;
                else
                    frames[j]->i_type = X264_TYPE_P;
            }
        }	//end of b_adaptive = X264_B_ADAPT_FAST
其中不使用自适应B帧(X264_B_ADAPT_NONE)部分:
/*
	过程:
		1.	若没有B帧的坑位了则将AUTO/B改P
		2.	若有B帧坑位,则
			·	后面一帧为B,当前设P
			·	后面一帧为P,当前设B
		3.	更新B帧的坑位
*/
else	//b_adaptive=0,即不使用自适应B帧
        {
            int num_bframes = h->param.i_bframe;
            for( int j = 1; j < num_frames; j++ )
            {
                if( !num_bframes )	//没有B帧坑位了
                {
                    if( IS_X264_TYPE_AUTO_OR_B( frames[j]->i_type ) )
                        frames[j]->i_type = X264_TYPE_P;	//设置为P
                }
                else if( frames[j]->i_type == X264_TYPE_AUTO )	//有B帧坑位
                {
                    if( IS_X264_TYPE_B( frames[j+1]->i_type ) )	//后面是B则当前设P,同样为什么?
                        frames[j]->i_type = X264_TYPE_P;
                    else										//后面是P则当前设B
                        frames[j]->i_type = X264_TYPE_B;
                }

				//更新B帧的坑位
                if( IS_X264_TYPE_B( frames[j]->i_type ) )	//B帧
                    num_bframes--;							//占一个坑位
                else
                    num_bframes = h->param.i_bframe;		//更新坑位
            }
        }	//end of b_adaptive = X264_B_ADAPT_NONE
 类似资料: