x264 struct 学习

x264_t结构体维护着CODEC的诸多重要信息

其中成员frames是一个指示和控制帧编码过程的结构。其中current是已经准备就绪可以编码的帧，其类型已经确定；next是尚未确定类型的帧；unused用于回收不使用的frame结构体以备今后再次使用。

structx264_t

{

unsigned int cpu;

struct

{

int i_nal; //定义一个网络提取层//

x264_nal_t nal[3];

int i_bitstream;

uint8_t *p_bitstream;

bs_t bs;

} out;

x264_param_t param;

int i_frame; //一个指示和控制帧编码过程的结构

int i_poc;

int i_frame_offset;

int i_frame_num;

int i_poc_msb;

int i_poc_lsb;

x264_sps_t sps_array[32];

x264_sps_t *sps;

x264_pps_t pps_array[256];

x264_pps_t *pps;

int i_idr_pic_id;

x264_slice_header_t sh;

x264_cabac_t cabac;

x264_picture_t *picture;

x264_frame_t *bframe_current[X264_BFRAME_MAX];

x264_frame_t *frame_next[X264_BFRAME_MAX+1]; 这个是定义下一个帧,但不一定是B帧

x264_frame_t *frame_unused[X264_BFRAME_MAX+1];

x264_frame_t *fdec;

struct

{

DECLARE_ALIGNED( int, luma16×16_dc[16], 16 );

DECLARE_ALIGNED( int, chroma_dc[2][4], 16 );//色度

struct

{

DECLARE_ALIGNED( int, residual_ac[15], 16 );

DECLARE_ALIGNED( int, luma4×4[16], 16 );

} block[16+8];16*8的宏快的取用情况

} dct;离散余弦变换

struct

{

int i_mb_stride;

int i_mb_x;

int i_mb_y;

int i_mb_xy;

unsigned int i_neighbour;

int8_t *type;

int8_t *qp;

int16_t*cbp; ？

int8_t (*intra4×4_pred_mode)[7];

uint8_t (*non_zero_count)[16+4+4];

int8_t *chroma_pred_mode;

int16_t(*mv[2])[2];

int16_t(*mvd[2])[2];

int8_t *ref[2];

int i_type;

int i_partition;

int i_sub_partition[4];

int i_cbp_luma;

int i_cbp_chroma;

int i_intra16×16_pred_mode;

int i_chroma_pred_mode;

struct

{

uint8_t *p_img[3];

int i_img[3];

uint8_t *p_fdec[3];

uint8_t*p_fref[2][16][3];

int i_fdec[3];

} pic;

struct

{

int intra4×4_pred_mode[X264_SCAN8_SIZE];色度预测模式

int non_zero_count[X264_SCAN8_SIZE];

int8_t ref[2][X264_SCAN8_SIZE];

int16_t mv[2][X264_SCAN8_SIZE][2];

int16_t mvd[2][X264_SCAN8_SIZE][2];运动矢量与真实值之间的差

} cache;

int i_last_qp;

int i_last_dqp;

} mb;宏快的预测以及运动补偿和运动估计

x264_frame_t *freference[16+1];

int i_ref0; 参考帧列表一（定义为lsit0意为向前预测编码）

x264_frame_t *fref0[16];

int i_ref1;参考帧列表二（定义为lsit1意为向后预测编码）

x264_frame_t *fref1[16];

x264_ratecontrol_t *rc;

struct

{

int i_slice_count[5];

int i_slice_size[5];

float f_psnr_y[5];

float f_psnr_u[5];

float f_psnr_v[5];

int i_mb_count[5][18];

} stat;声明每一片的参数变量

x264_predict_t predict_16×16[4+3];

x264_predict_t predict_8×8[4+3];

x264_predict_t predict_4×4[9+3];

x264_pixel_function_t pixf;

x264_mc_function_t mc[2];

x264_dct_function_t dctf;

x264_vlc_table_t *x264_coeff_token_lookup[5];

x264_vlc_table_t *x264_level_prefix_lookup;

x264_vlc_table_t *x264_total_zeros_lookup[15];

x264_vlc_table_t *x264_total_zeros_dc_lookup[3];

x264_vlc_table_t *x264_run_before_lookup[7];

};

#endif

给出编码过程中的各个参数的初始化

x264_param_t

typedef struct

{

unsigned int cpu;

int i_width;

int i_height;

struct

{

int i_sar_height;

int i_sar_width;

} vui;？

float f_fps;

int i_frame_reference;

int i_idrframe;

int i_iframe;

int i_bframe;

int b_deblocking_filter;

int b_cabac;

int i_cabac_init_idc;给出熵编码初始化时表格的选择

int i_qp_constant;

int i_bitrate;

struct

{

unsigned intintra;

unsigned intinter;

} analyse;

}
x264_param_t给出编码过程中的各个参数的初始化

sps序列参数集

typedef struct

{

int i_id;

int i_profile_idc;指明所用profile

int i_level_idc; 指明所用level

int b_constraint_set0;

int b_constraint_set1;

int b_constraint_set2; 当他们值等于1时，表示必须遵守某些制约条件

int i_log2_max_frame_num;表示图像解码顺序的最大取值

inti_poc_type;

inti_log2_max_poc_lsb;

intb_delta_pic_order_always_zero;其值等于一时句法元素delta_pic_order_cnt[0]和delta_pic_order_cnt[1]不再片头出现，且他们的默认值都为0。为0时上述则出现。

int i_offset_for_non_ref_pic;用来计算非参考帧或场的picture order count,其值应在[-2e31,2e31-1]

int i_offset_for_top_to_bottom_field; 用来计算帧的底场的picture order count其值应在[-2e31,2e31-1]

int i_num_ref_frames_in_poc_cycle;用来解码picture order count取值应在[0，255]之间

int i_offset_for_ref_frame[256]; 当picture order counttype=1时用来解码poc，这句语法对循环num_ref_frames_in_poc_cycle中的每一个元素指定了一个偏移

int i_num_ref_frames;指定参考帧队列的最大长度 h264规定最多可为16个参考帧

int b_gaps_in_frame_num_value_allowed;值为1时表示允许句法frame_num可以不连续

int i_mb_width;

int i_mb_height;

int b_frame_mbs_only;

int b_mb_adaptive_frame_field;

   int b_direct8×8_inference;指明b片的直接和skip模式下运动矢量的预测方法

   int b_crop;

struct

{

int i_left;

int i_right;

int i_top;

int i_bottom;