LiveRTMP之H264 SPS解析宽高不正确导致播放时画面拉伸的问题修复（五）

 

LiveRTMP

LiveRTMP是将H264流以及AAC流以RTMP协议推送到RTMP服务器上进行直播。LiveRTMP推送库中会从H264流中提取中SPS、PPS进行解析，开发的时候遇到过有些SPS解析有误，获取到的宽高不正确，导致播放的时候画面被拉长的问题。下面给出一份比较完善的SPS解析

##SPS解析
定义sps结构

typedef struct
{
    int i_id;

    int i_profile_idc;
    int i_level_idc;

    int b_constraint_set0;
    int b_constraint_set1;
    int b_constraint_set2;

    int i_chroma_format_idc;
    int i_log2_max_frame_num;

    int i_poc_type;
    /* poc 0 */
    int i_log2_max_poc_lsb;
    /* poc 1 */
    int b_delta_pic_order_always_zero;
    int i_offset_for_non_ref_pic;
    int i_offset_for_top_to_bottom_field;
    int i_num_ref_frames_in_poc_cycle;
    int i_offset_for_ref_frame[256];

    int i_num_ref_frames;
    int b_gaps_in_frame_num_value_allowed;
    int i_mb_width;
    int i_mb_height;
    int b_frame_mbs_only;
    int b_mb_adaptive_frame_field;
    int b_direct8x8_inference;

    int b_crop;
    struct
    {
        int i_left;
        int i_right;
        int i_top;
        int i_bottom;
    } crop;

    int b_vui;
    struct
    {
        int b_aspect_ratio_info_present;
        int i_sar_width;
        int i_sar_height;
        
        int b_overscan_info_present;
        int b_overscan_info;

        int b_signal_type_present;
        int i_vidformat;
        int b_fullrange;
        int b_color_description_present;
        int i_colorprim;
        int i_transfer;
        int i_colmatrix;

        int b_chroma_loc_info_present;
        int i_chroma_loc_top;
        int i_chroma_loc_bottom;

        int b_timing_info_present;
        int i_num_units_in_tick;
        int i_time_scale;
        int b_fixed_frame_rate;

    	int nal_hrd_parameters_present_flag;
    	int vcl_hrd_parameters_present_flag;
    	int pic_struct_present_flag;

        int b_bitstream_restriction;
        int b_motion_vectors_over_pic_boundaries;
        int i_max_bytes_per_pic_denom;
        int i_max_bits_per_mb_denom;
        int i_log2_max_mv_length_horizontal;
        int i_log2_max_mv_length_vertical;
        int i_num_reorder_frames;
        int i_max_dec_frame_buffering;

        /* FIXME to complete */
    } vui;

    int b_qpprime_y_zero_transform_bypass;

    int scaling_matrix_present;
    uint8_t scaling_matrix4[6][16];
    uint8_t scaling_matrix8[6][64];
} h264_sps_t;

解析SPS数据，得到h264_sps_t结构的数据

/* return -1 if invalid, else the id */
int h264_sps_read( unsigned char *nal, int nal_len, h264_sps_t *sps)
{
    int i_profile_idc;
    int i_level_idc;

    int b_constraint_set0;
    int b_constraint_set1;
    int b_constraint_set2;

    int id;
    bs_t bs;
    bs_t *s = &bs;

    sps->scaling_matrix_present = 0;

    bs_init( &bs, nal+1, nal_len-1 );

    //P264_TRACE_ADDRESS();
    i_profile_idc     = bs_read( s, 8 );
    //_TRACE2("SPS: profile_idc = %d\n", i_profile_idc);
    b_constraint_set0 = bs_read( s, 1 );
    b_constraint_set1 = bs_read( s, 1 );
    b_constraint_set2 = bs_read( s, 1 );

    bs_skip( s, 5 );    /* reserved */
    //P264_TRACE_ADDRESS();
    i_level_idc       = bs_read( s, 8 );
    //_TRACE2("SPS: level_idc = %d\n", i_level_idc);


    id = bs_read_ue( s );
    if( bs_eof( s ) || id >= 32 )
    {
        /* the sps is invalid, no need to corrupt sps_array[0] */
        return -1;
    }

    sps->i_id = id;

    /* put pack parsed value */
    sps->i_profile_idc     = i_profile_idc;
    sps->i_level_idc       = i_level_idc;
    sps->b_constraint_set0 = b_constraint_set0;
    sps->b_constraint_set1 = b_constraint_set1;
    sps->b_constraint_set2 = b_constraint_set2;

    if(sps->i_profile_idc >= 100){ //high profile
        sps->i_chroma_format_idc= bs_read_ue( s );
        if(sps->i_chroma_format_idc >= 32 )
            return -1;
        if(sps->i_chroma_format_idc == 3)
            bs_read( s, 1 );//residual_color_transform_flag
        //sps->bit_depth_luma   = get_ue_golomb(&s->gb) + 8;
        //sps->bit_depth_chroma = get_ue_golomb(&s->gb) + 8;
        //sps->transform_bypass = get_bits1(&s->gb);

		/* bit_depth_luma_minus8 */
        bs_read_ue( s );
		/* bit_depth_chroma_minus8 */
        bs_read_ue( s );
		/* qpprime_y_zero_transform_bypass_flag */
		bs_skip( s, 1 );
		/* seq_scaling_matrix_present_flag */
        int seq_scaling_matrix_present_flag = bs_read( s, 1 );
        //decode_scaling_matrices(s, sps, NULL, 1, sps->scaling_matrix4, sps->scaling_matrix8);

			if (seq_scaling_matrix_present_flag)
			{
				for( int i = 0; i < ((3 != sps->i_chroma_format_idc) ? 8 : 12); i++ )
				{
					/* seq_scaling_list_present_flag[i] */
					seq_scaling_matrix_present_flag = bs_read( s, 1 );
					if( !seq_scaling_matrix_present_flag )
						continue;
					const int i_size_of_scaling_list = (i < 6 ) ? 16 : 64;
					/* scaling_list (...) */
					int i_lastscale = 8;
					int i_nextscale = 8;
					for( int j = 0; j < i_size_of_scaling_list; j++ )
					{
						if( i_nextscale != 0 )
						{
							/* delta_scale */
							seq_scaling_matrix_present_flag = bs_read_se( s );
							i_nextscale = ( i_lastscale + seq_scaling_matrix_present_flag + 256 ) % 256;
							/* useDefaultScalingMatrixFlag = ... */
						}
						/* scalinglist[j] */
						i_lastscale = ( i_nextscale == 0 ) ? i_lastscale : i_nextscale;
					}
				}
			}
    }

    //sps->i_log2_max_frame_num = bs_read_ue( s ) + 4;
	//sps->i_log2_max_frame_num = 2^(bs_read_ue( s ) + 4);

	sps->i_log2_max_frame_num = bs_read_ue( s );


    sps->i_poc_type = bs_read_ue( s );
    if( sps->i_poc_type == 0 )
    {
        sps->i_log2_max_poc_lsb = bs_read_ue( s ) + 4;
    }
    else if( sps->i_poc_type == 1 )
    {
        int i;
        sps->b_delta_pic_order_always_zero = bs_read( s, 1 );
        sps->i_offset_for_non_ref_pic = bs_read_se( s );
        sps->i_offset_for_top_to_bottom_field = bs_read_se( s );
        sps->i_num_ref_frames_in_poc_cycle = bs_read_ue( s );
        if( sps->i_num_ref_frames_in_poc_cycle > 256 )
        {
            /* FIXME what to do */
            sps->i_num_ref_frames_in_poc_cycle = 256;
        }
        for( i = 0; i < sps->i_num_ref_frames_in_poc_cycle; i++ )
        {
            sps->i_offset_for_ref_frame[i] = bs_read_se( s );
        }
    }
    else if( sps->i_poc_type > 2 )
    {
        goto error;
    }

    sps->i_num_ref_frames = bs_read_ue( s );
    //_TRACE2("SPS: num_ref_frames = %d\n", sps->i_num_ref_frames);
    sps->b_gaps_in_frame_num_value_allowed = bs_read( s, 1 );
    sps->i_mb_width = bs_read_ue( s ) + 1;
    //_TRACE2("SPS: mb_width = %d\n", sps->i_mb_width);
    sps->i_mb_height= bs_read_ue( s ) + 1;
    //_TRACE2("SPS: mb_height = %d\n", sps->i_mb_height);
    sps->b_frame_mbs_only = bs_read( s, 1 );
    if( !sps->b_frame_mbs_only )
    {
        sps->b_mb_adaptive_frame_field = bs_read( s, 1 );
    }
    else
    {
        sps->b_mb_adaptive_frame_field = 0;
    }
    sps->b_direct8x8_inference = bs_read( s, 1 );

    sps->b_crop = bs_read( s, 1 );
    if( sps->b_crop )
    {
        sps->crop.i_left  = bs_read_ue( s );
        sps->crop.i_right = bs_read_ue( s );
        sps->crop.i_top   = bs_read_ue( s );
        sps->crop.i_bottom= bs_read_ue( s );
    }
    else
    {
        sps->crop.i_left  = 0;
        sps->crop.i_right = 0;
        sps->crop.i_top   = 0;
        sps->crop.i_bottom= 0;
    }

    sps->b_vui = bs_read( s, 1 );
    if( sps->b_vui )
    {
        /* FIXME */
        //_TRACE2( "decode vui %d\n", bs_pos(s) );
        decode_vui_parameters(s, sps);
    }

    if( bs_eof( s ) )
    {
        /* no rbsp trailing */
        //_TRACE2( "incomplete SPS\n" );
		sps->i_id = -1;
		return -1000;
    }

    return id;

error:
    /* invalidate this sps */
    sps->i_id = -1;
    return -1;
}