下面将开始分析x264编码的核心部分:x264_encoder_encode函数。 首先先看encode_frame函数:
static int encode_frame( x264_t *h, hnd_t hout, x264_picture_t *pic, int64_t *last_dts )
{
x264_picture_t pic_out; //编码后的图片
x264_nal_t *nal;
int i_nal;
int i_frame_size = 0; //帧的大小
//将待编码的一帧图片作为参数传进去,进行编码,再将编码后的图片传回来
i_frame_size = x264_encoder_encode( h, &nal, &i_nal, pic, &pic_out );
FAIL_IF_ERROR( i_frame_size < 0, "x264_encoder_encode failed\n" );
if( i_frame_size )
{
//将编码后的一帧图片输出
i_frame_size = cli_output.write_frame( hout, nal[0].p_payload, i_frame_size, &pic_out );
*last_dts = pic_out.i_dts; //跟新dts(编码顺序)
}
return i_frame_size;
}
/****************************************************************************
* x264_encoder_encode:
* XXX: i_poc : is the poc of the current given picture
* i_frame : is the number of the frame being coded
* ex: type frame poc
* I 0 2*0
* P 1 2*3
* B 2 2*1
* B 3 2*2
* P 4 2*6
* B 5 2*4
* B 6 2*5
****************************************************************************/
int x264_encoder_encode( x264_t *h,
x264_nal_t **pp_nal, int *pi_nal,
x264_picture_t *pic_in,
x264_picture_t *pic_out )
{
x264_t *thread_current, *thread_prev, *thread_oldest;
int i_nal_type, i_nal_ref_idc, i_global_qp;
int overhead = NALU_OVERHEAD;
#if HAVE_OPENCL
if( h->opencl.b_fatal_error )
return -1;
#endif
if( h->i_thread_frames > 1 )
{
thread_prev = h->thread[ h->i_thread_phase ];
h->i_thread_phase = (h->i_thread_phase + 1) % h->i_thread_frames;
thread_current = h->thread[ h->i_thread_phase ];
thread_oldest = h->thread[ (h->i_thread_phase + 1) % h->i_thread_frames ];
x264_thread_sync_context( thread_current, thread_prev );
x264_thread_sync_ratecontrol( thread_current, thread_prev, thread_oldest );
h = thread_current;
}
else
{
thread_current =
thread_oldest = h;
}
h->i_cpb_delay_pir_offset = h->i_cpb_delay_pir_offset_next;
/* no data out */
*pi_nal = 0;
*pp_nal = NULL;
/* ------------------- Setup new frame from picture -------------------- */
if( pic_in != NULL )
{
/* 1: Copy the picture to a frame and move it to a buffer */
//指针指向未编码的一帧空间
x264_frame_t *fenc = x264_frame_pop_unused( h, 0 );
if( !fenc )
return -1;
//将图片的信息拷贝到fenc中
if( x264_frame_copy_picture( h, fenc, pic_in ) < 0 )
return -1;
//如果宽高不是16的倍数,进行扩展
if( h->param.i_width != 16 * h->mb.i_mb_width ||
h->param.i_height != 16 * h->mb.i_mb_height )
x264_frame_expand_border_mod16( h, fenc );
//记录帧号
fenc->i_frame = h->frames.i_input++;
if( fenc->i_frame == 0 )
h->frames.i_first_pts = fenc->i_pts;
if( h->frames.i_bframe_delay && fenc->i_frame == h->frames.i_bframe_delay )
h->frames.i_bframe_delay_time = fenc->i_pts - h->frames.i_first_pts;
if( h->param.b_vfr_input && fenc->i_pts <= h->frames.i_largest_pts )
x264_log( h, X264_LOG_WARNING, "non-strictly-monotonic PTS\n" );
h->frames.i_second_largest_pts = h->frames.i_largest_pts;
h->frames.i_largest_pts = fenc->i_pts;
if( (fenc->i_pic_struct < PIC_STRUCT_AUTO) || (fenc->i_pic_struct > PIC_STRUCT_TRIPLE) )
fenc->i_pic_struct = PIC_STRUCT_AUTO;
if( fenc->i_pic_struct == PIC_STRUCT_AUTO )
{
#if HAVE_INTERLACED
int b_interlaced = fenc->param ? fenc->param->b_interlaced : h->param.b_interlaced;
#else
int b_interlaced = 0;
#endif
if( b_interlaced )
{
int b_tff = fenc->param ? fenc->param->b_tff : h->param.b_tff;
fenc->i_pic_struct = b_tff ? PIC_STRUCT_TOP_BOTTOM : PIC_STRUCT_BOTTOM_TOP;
}
else
fenc->i_pic_struct = PIC_STRUCT_PROGRESSIVE;
}
if( h->param.rc.b_mb_tree && h->param.rc.b_stat_read )
{
if( x264_macroblock_tree_read( h, fenc, pic_in->prop.quant_offsets ) )
return -1;
}
else
x264_stack_align( x264_adaptive_quant_frame, h, fenc, pic_in->prop.quant_offsets );
if( pic_in->prop.quant_offsets_free )
pic_in->prop.quant_offsets_free( pic_in->prop.quant_offsets );
if( h->frames.b_have_lowres )
x264_frame_init_lowres( h, fenc );
/* 2: Place the frame into the queue for its slice type decision */
//进行lookahead操作(对slice类型进行判定,并进行排序)
x264_lookahead_put_frame( h, fenc );
if( h->frames.i_input <= h->frames.i_delay + 1 - h->i_thread_frames )
{
/* Nothing yet to encode, waiting for filling of buffers */
pic_out->i_type = X264_TYPE_AUTO;
return 0;
}
}
else
{
/* signal kills for lookahead thread */
x264_pthread_mutex_lock( &h->lookahead->ifbuf.mutex );
h->lookahead->b_exit_thread = 1;
x264_pthread_cond_broadcast( &h->lookahead->ifbuf.cv_fill );
x264_pthread_mutex_unlock( &h->lookahead->ifbuf.mutex );
}
h->i_frame++;
/* 3: The picture is analyzed in the lookahead */
//从lookahead中获得分析完的帧
if( !h->frames.current[0] )
x264_lookahead_get_frames( h );
if( !h->frames.current[0] && x264_lookahead_is_empty( h ) )
return x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
/* ------------------- Get frame to be encoded ------------------------- */
//开始编码
/* 4: get picture to encode */
//取出第一帧作为编码帧,后面的帧依次向前移
h->fenc = x264_frame_shift( h->frames.current );
/* If applicable, wait for previous frame reconstruction to finish */
if( h->param.b_sliced_threads )
if( x264_threadpool_wait_all( h ) < 0 )
return -1;
if( h->i_frame == h->i_thread_frames - 1 )
h->i_reordered_pts_delay = h->fenc->i_reordered_pts;
if( h->reconfig )
{
x264_encoder_reconfig_apply( h, &h->reconfig_h->param );
h->reconfig = 0;
}
if( h->fenc->param )
{
x264_encoder_reconfig_apply( h, h->fenc->param );
if( h->fenc->param->param_free )
{
h->fenc->param->param_free( h->fenc->param );
h->fenc->param = NULL;
}
}
// ok to call this before encoding any frames, since the initial values of fdec have b_kept_as_ref=0
//更新参考帧,将重建帧加入到参考队列中
if( x264_reference_update( h ) )
return -1;
h->fdec->i_lines_completed = -1;
if( !IS_X264_TYPE_I( h->fenc->i_type ) )
{
int valid_refs_left = 0;
for( int i = 0; h->frames.reference[i]; i++ )
if( !h->frames.reference[i]->b_corrupt )
valid_refs_left++;
/* No valid reference frames left: force an IDR. */
if( !valid_refs_left )
{
h->fenc->b_keyframe = 1;
h->fenc->i_type = X264_TYPE_IDR;
}
}
if( h->fenc->b_keyframe )
{
h->frames.i_last_keyframe = h->fenc->i_frame;
if( h->fenc->i_type == X264_TYPE_IDR )
{
h->i_frame_num = 0;
h->frames.i_last_idr = h->fenc->i_frame;
}
}
h->sh.i_mmco_command_count =
h->sh.i_mmco_remove_from_end = 0;
h->b_ref_reorder[0] =
h->b_ref_reorder[1] = 0;
h->fdec->i_poc =
h->fenc->i_poc = 2 * ( h->fenc->i_frame - X264_MAX( h->frames.i_last_idr, 0 ) );
/* ------------------- Setup frame context ----------------------------- */
/* 5: Init data dependent of frame type */
//根据帧的类型,进行初始化
if( h->fenc->i_type == X264_TYPE_IDR )
{
/* reset ref pictures */
i_nal_type = NAL_SLICE_IDR;
i_nal_ref_idc = NAL_PRIORITY_HIGHEST;
h->sh.i_type = SLICE_TYPE_I;
x264_reference_reset( h );
h->frames.i_poc_last_open_gop = -1;
}
else if( h->fenc->i_type == X264_TYPE_I )
{
i_nal_type = NAL_SLICE;
i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
h->sh.i_type = SLICE_TYPE_I;
x264_reference_hierarchy_reset( h );
if( h->param.b_open_gop )
h->frames.i_poc_last_open_gop = h->fenc->b_keyframe ? h->fenc->i_poc : -1;
}
else if( h->fenc->i_type == X264_TYPE_P )
{
i_nal_type = NAL_SLICE;
i_nal_ref_idc = NAL_PRIORITY_HIGH; /* Not completely true but for now it is (as all I/P are kept as ref)*/
h->sh.i_type = SLICE_TYPE_P;
x264_reference_hierarchy_reset( h );
h->frames.i_poc_last_open_gop = -1;
}
else if( h->fenc->i_type == X264_TYPE_BREF )
{
i_nal_type = NAL_SLICE;
i_nal_ref_idc = h->param.i_bframe_pyramid == X264_B_PYRAMID_STRICT ? NAL_PRIORITY_LOW : NAL_PRIORITY_HIGH;
h->sh.i_type = SLICE_TYPE_B;
x264_reference_hierarchy_reset( h );
}
else /* B frame */
{
i_nal_type = NAL_SLICE;
i_nal_ref_idc = NAL_PRIORITY_DISPOSABLE;
h->sh.i_type = SLICE_TYPE_B;
}
h->fdec->i_type = h->fenc->i_type;
h->fdec->i_frame = h->fenc->i_frame;
h->fenc->b_kept_as_ref =
h->fdec->b_kept_as_ref = i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE && h->param.i_keyint_max > 1;
h->fdec->mb_info = h->fenc->mb_info;
h->fdec->mb_info_free = h->fenc->mb_info_free;
h->fenc->mb_info = NULL;
h->fenc->mb_info_free = NULL;
h->fdec->i_pts = h->fenc->i_pts;
if( h->frames.i_bframe_delay )
{
int64_t *prev_reordered_pts = thread_current->frames.i_prev_reordered_pts;
h->fdec->i_dts = h->i_frame > h->frames.i_bframe_delay
? prev_reordered_pts[ (h->i_frame - h->frames.i_bframe_delay) % h->frames.i_bframe_delay ]
: h->fenc->i_reordered_pts - h->frames.i_bframe_delay_time;
prev_reordered_pts[ h->i_frame % h->frames.i_bframe_delay ] = h->fenc->i_reordered_pts;
}
else
h->fdec->i_dts = h->fenc->i_reordered_pts;
if( h->fenc->i_type == X264_TYPE_IDR )
h->i_last_idr_pts = h->fdec->i_pts;
/* ------------------- Init ----------------------------- */
/* build ref list 0/1 */
//重建参考帧列表list0和list1,即将参考帧列表中的参考帧分为前向参考帧和后向参考帧,并根据POC进行参考帧排序。
//参考帧列表list0按降序排列,参考帧列表list1按升序排列
x264_reference_build_list( h, h->fdec->i_poc );
/* ---------------------- Write the bitstream -------------------------- */
/* Init bitstream context */
//初始化比特流
if( h->param.b_sliced_threads )
{
for( int i = 0; i < h->param.i_threads; i++ )
{
bs_init( &h->thread[i]->out.bs, h->thread[i]->out.p_bitstream, h->thread[i]->out.i_bitstream );
h->thread[i]->out.i_nal = 0;
}
}
else
{
bs_init( &h->out.bs, h->out.p_bitstream, h->out.i_bitstream );
h->out.i_nal = 0;
}
if( h->param.b_aud )
{
int pic_type;
if( h->sh.i_type == SLICE_TYPE_I )
pic_type = 0;
else if( h->sh.i_type == SLICE_TYPE_P )
pic_type = 1;
else if( h->sh.i_type == SLICE_TYPE_B )
pic_type = 2;
else
pic_type = 7;
x264_nal_start( h, NAL_AUD, NAL_PRIORITY_DISPOSABLE );
bs_write( &h->out.bs, 3, pic_type );
bs_rbsp_trailing( &h->out.bs );
if( x264_nal_end( h ) )
return -1;
overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
}
h->i_nal_type = i_nal_type;
h->i_nal_ref_idc = i_nal_ref_idc;
if( h->param.b_intra_refresh )
{
if( IS_X264_TYPE_I( h->fenc->i_type ) )
{
h->fdec->i_frames_since_pir = 0;
h->b_queued_intra_refresh = 0;
/* PIR is currently only supported with ref == 1, so any intra frame effectively refreshes
* the whole frame and counts as an intra refresh. */
h->fdec->f_pir_position = h->mb.i_mb_width;
}
else if( h->fenc->i_type == X264_TYPE_P )
{
int pocdiff = (h->fdec->i_poc - h->fref[0][0]->i_poc)/2;
float increment = X264_MAX( ((float)h->mb.i_mb_width-1) / h->param.i_keyint_max, 1 );
h->fdec->f_pir_position = h->fref[0][0]->f_pir_position;
h->fdec->i_frames_since_pir = h->fref[0][0]->i_frames_since_pir + pocdiff;
if( h->fdec->i_frames_since_pir >= h->param.i_keyint_max ||
(h->b_queued_intra_refresh && h->fdec->f_pir_position + 0.5 >= h->mb.i_mb_width) )
{
h->fdec->f_pir_position = 0;
h->fdec->i_frames_since_pir = 0;
h->b_queued_intra_refresh = 0;
h->fenc->b_keyframe = 1;
}
h->fdec->i_pir_start_col = h->fdec->f_pir_position+0.5;
h->fdec->f_pir_position += increment * pocdiff;
h->fdec->i_pir_end_col = h->fdec->f_pir_position+0.5;
/* If our intra refresh has reached the right side of the frame, we're done. */
if( h->fdec->i_pir_end_col >= h->mb.i_mb_width - 1 )
{
h->fdec->f_pir_position = h->mb.i_mb_width;
h->fdec->i_pir_end_col = h->mb.i_mb_width - 1;
}
}
}
if( h->fenc->b_keyframe )
{
/* Write SPS and PPS */
//写入SPS和PPS
if( h->param.b_repeat_headers )
{
/* generate sequence parameters */
x264_nal_start( h, NAL_SPS, NAL_PRIORITY_HIGHEST ); //NAL单元首地址被赋值,设置NAL优先权和类型
x264_sps_write( &h->out.bs, h->sps ); //写入SPS
if( x264_nal_end( h ) ) //结束NAL,整理NAL单元,更新指针,如果空间不够,将重新分配空间
return -1;
/* Pad AUD/SPS to 256 bytes like Panasonic */
if( h->param.i_avcintra_class )
h->out.nal[h->out.i_nal-1].i_padding = 256 - bs_pos( &h->out.bs ) / 8 - 2*NALU_OVERHEAD;
overhead += h->out.nal[h->out.i_nal-1].i_payload + h->out.nal[h->out.i_nal-1].i_padding + NALU_OVERHEAD;
/* generate picture parameters */
x264_nal_start( h, NAL_PPS, NAL_PRIORITY_HIGHEST );
x264_pps_write( &h->out.bs, h->sps, h->pps ); //写入PPS
if( x264_nal_end( h ) )
return -1;
if( h->param.i_avcintra_class )
h->out.nal[h->out.i_nal-1].i_padding = 256 - h->out.nal[h->out.i_nal-1].i_payload - NALU_OVERHEAD;
overhead += h->out.nal[h->out.i_nal-1].i_payload + h->out.nal[h->out.i_nal-1].i_padding + NALU_OVERHEAD;
}
/* when frame threading is used, buffering period sei is written in x264_encoder_frame_end */
//写入SEI
if( h->i_thread_frames == 1 && h->sps->vui.b_nal_hrd_parameters_present )
{
x264_hrd_fullness( h );
x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
x264_sei_buffering_period_write( h, &h->out.bs );
if( x264_nal_end( h ) )
return -1;
overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
}
}
/* write extra sei */
for( int i = 0; i < h->fenc->extra_sei.num_payloads; i++ )
{
x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
x264_sei_write( &h->out.bs, h->fenc->extra_sei.payloads[i].payload, h->fenc->extra_sei.payloads[i].payload_size,
h->fenc->extra_sei.payloads[i].payload_type );
if( x264_nal_end( h ) )
return -1;
overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
if( h->fenc->extra_sei.sei_free )
{
h->fenc->extra_sei.sei_free( h->fenc->extra_sei.payloads[i].payload );
h->fenc->extra_sei.payloads[i].payload = NULL;
}
}
if( h->fenc->extra_sei.sei_free )
{
h->fenc->extra_sei.sei_free( h->fenc->extra_sei.payloads );
h->fenc->extra_sei.payloads = NULL;
h->fenc->extra_sei.sei_free = NULL;
}
if( h->fenc->b_keyframe )
{
/* Avid's decoder strictly wants two SEIs for AVC-Intra so we can't insert the x264 SEI */
if( h->param.b_repeat_headers && h->fenc->i_frame == 0 && !h->param.i_avcintra_class )
{
/* identify ourself */
x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
if( x264_sei_version_write( h, &h->out.bs ) )
return -1;
if( x264_nal_end( h ) )
return -1;
overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
}
if( h->fenc->i_type != X264_TYPE_IDR )
{
int time_to_recovery = h->param.b_open_gop ? 0 : X264_MIN( h->mb.i_mb_width - 1, h->param.i_keyint_max ) + h->param.i_bframe - 1;
x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
x264_sei_recovery_point_write( h, &h->out.bs, time_to_recovery );
if( x264_nal_end( h ) )
return -1;
overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
}
if( h->param.i_frame_packing >= 0 )
{
x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
x264_sei_frame_packing_write( h, &h->out.bs );
if( x264_nal_end( h ) )
return -1;
overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
}
}
/* generate sei pic timing */
if( h->sps->vui.b_pic_struct_present || h->sps->vui.b_nal_hrd_parameters_present )
{
x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
x264_sei_pic_timing_write( h, &h->out.bs );
if( x264_nal_end( h ) )
return -1;
overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
}
/* As required by Blu-ray. */
if( !IS_X264_TYPE_B( h->fenc->i_type ) && h->b_sh_backup )
{
h->b_sh_backup = 0;
x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
x264_sei_dec_ref_pic_marking_write( h, &h->out.bs );
if( x264_nal_end( h ) )
return -1;
overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
}
if( h->fenc->b_keyframe && h->param.b_intra_refresh )
h->i_cpb_delay_pir_offset_next = h->fenc->i_cpb_delay;
/* Filler space: 10 or 18 SEIs' worth of space, depending on resolution */
if( h->param.i_avcintra_class )
{
/* Write an empty filler NAL to mimic the AUD in the P2 format*/
x264_nal_start( h, NAL_FILLER, NAL_PRIORITY_DISPOSABLE );
x264_filler_write( h, &h->out.bs, 0 );
if( x264_nal_end( h ) )
return -1;
overhead += h->out.nal[h->out.i_nal-1].i_payload + NALU_OVERHEAD;
/* All lengths are magic lengths that decoders expect to see */
/* "UMID" SEI */
x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
if( x264_sei_avcintra_umid_write( h, &h->out.bs ) < 0 )
return -1;
if( x264_nal_end( h ) )
return -1;
overhead += h->out.nal[h->out.i_nal-1].i_payload + SEI_OVERHEAD;
int unpadded_len;
int total_len;
if( h->param.i_height == 1080 )
{
unpadded_len = 5780;
total_len = 17*512;
}
else
{
unpadded_len = 2900;
total_len = 9*512;
}
/* "VANC" SEI */
x264_nal_start( h, NAL_SEI, NAL_PRIORITY_DISPOSABLE );
if( x264_sei_avcintra_vanc_write( h, &h->out.bs, unpadded_len ) < 0 )
return -1;
if( x264_nal_end( h ) )
return -1;
h->out.nal[h->out.i_nal-1].i_padding = total_len - h->out.nal[h->out.i_nal-1].i_payload - SEI_OVERHEAD;
overhead += h->out.nal[h->out.i_nal-1].i_payload + h->out.nal[h->out.i_nal-1].i_padding + SEI_OVERHEAD;
}
/* Init the rate control */
//比特率控制初始化
/* FIXME: Include slice header bit cost. */
x264_ratecontrol_start( h, h->fenc->i_qpplus1, overhead*8 );
i_global_qp = x264_ratecontrol_qp( h );
pic_out->i_qpplus1 =
h->fdec->i_qpplus1 = i_global_qp + 1;
if( h->param.rc.b_stat_read && h->sh.i_type != SLICE_TYPE_I )
{
x264_reference_build_list_optimal( h );
x264_reference_check_reorder( h );
}
if( h->i_ref[0] )
h->fdec->i_poc_l0ref0 = h->fref[0][0]->i_poc;
/* ------------------------ Create slice header ----------------------- */
//初始化片,并写入片的头部数据
x264_slice_init( h, i_nal_type, i_global_qp );
/*------------------------- Weights -------------------------------------*/
if( h->sh.i_type == SLICE_TYPE_B )
x264_macroblock_bipred_init( h );
x264_weighted_pred_init( h );
if( i_nal_ref_idc != NAL_PRIORITY_DISPOSABLE )
h->i_frame_num++;
/* Write frame */
//下面将是片编码,其中包括宏块编码
h->i_threadslice_start = 0;
h->i_threadslice_end = h->mb.i_mb_height;
if( h->i_thread_frames > 1 )
{
x264_threadpool_run( h->threadpool, (void*)x264_slices_write, h );
h->b_thread_active = 1;
}
else if( h->param.b_sliced_threads )
{
if( x264_threaded_slices_write( h ) )
return -1;
}
else
if( (intptr_t)x264_slices_write( h ) )
return -1;
//编码后的收尾工作
return x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out );
}
至此,一帧的图像编码结束。