/*
* Generic DCT based hybrid video encoder
* Copyright (c) 2000, 2001, 2002 Fabrice Bellard.
+ * Copyright (c) 2002-2004 Michael Niedermayer
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
#define AVCODEC_MPEGVIDEO_H
#include "dsputil.h"
+#include "bitstream.h"
#define FRAME_SKIPED 100 ///< return value for header parsers if frame is not coded
enum OutputFormat {
FMT_MPEG1,
+ FMT_H261,
FMT_H263,
FMT_MJPEG,
FMT_H264,
#define MAX_FCODE 7
#define MAX_MV 2048
-#define MAX_PICTURE_COUNT 15
+#define MAX_THREADS 8
+
+#define MAX_PICTURE_COUNT 32
#define ME_MAP_SIZE 64
#define ME_MAP_SHIFT 3
#define SI_TYPE FF_SI_TYPE ///< Switching Intra
#define SP_TYPE FF_SP_TYPE ///< Switching Predicted
+#define MAX_MB_BYTES (30*16*16*3/8 + 120)
+
typedef struct Predictor{
double coeff;
double count;
*/
uint8_t *interpolated[3];
int16_t (*motion_val_base[2])[2];
- int8_t *ref_index[2];
uint32_t *mb_type_base;
+#define MB_TYPE_INTRA MB_TYPE_INTRA4x4 //default mb_type if theres just one type
#define IS_INTRA4x4(a) ((a)&MB_TYPE_INTRA4x4)
#define IS_INTRA16x16(a) ((a)&MB_TYPE_INTRA16x16)
#define IS_PCM(a) ((a)&MB_TYPE_INTRA_PCM)
int frame_num; ///< h264 frame_num
int pic_id; ///< h264 pic_num or long_term_pic_idx
int long_ref; ///< 1->long term reference 0->short term reference
+ int ref_poc[2][16]; ///< h264 POCs of the frames used as reference
+ int ref_count[2]; ///< number of entries in ref_poc
int mb_var_sum; ///< sum of MB variance for current frame
int mc_mb_var_sum; ///< motion compensated MB variance for current frame
* Motion estimation context.
*/
typedef struct MotionEstContext{
+ AVCodecContext *avctx;
int skip; ///< set if ME is skiped for the current MB
int co_located_mv[4][2]; ///< mv from last p frame for direct mode ME
int direct_basis_mv[4][2];
uint8_t *scratchpad; ///< data area for the me algo, so that the ME doesnt need to malloc/free
+ uint8_t *best_mb;
+ uint8_t *temp_mb[2];
+ uint8_t *temp;
+ int best_bits;
uint32_t *map; ///< map to avoid duplicate evaluations
uint32_t *score_map; ///< map to store the scores
int map_generation;
int penalty_factor;
int sub_penalty_factor;
int mb_penalty_factor;
+ int flags;
+ int sub_flags;
+ int mb_flags;
int pre_pass; ///< = 1 for the pre pass
int dia_size;
+ int xmin;
+ int xmax;
+ int ymin;
+ int ymax;
+ int pred_x;
+ int pred_y;
+ uint8_t *src[4][4];
+ uint8_t *ref[4][4];
+ int stride;
+ int uvstride;
+ /* temp variables for picture complexity calculation */
+ int mc_mb_var_sum_temp;
+ int mb_var_sum_temp;
+ int scene_change_score;
+/* cmp, chroma_cmp;*/
+ op_pixels_func (*hpel_put)[4];
+ op_pixels_func (*hpel_avg)[4];
+ qpel_mc_func (*qpel_put)[16];
+ qpel_mc_func (*qpel_avg)[16];
uint8_t (*mv_penalty)[MAX_MV*2+1]; ///< amount of bits needed to encode a MV
+ uint8_t *current_mv_penalty;
int (*sub_motion_search)(struct MpegEncContext * s,
int *mx_ptr, int *my_ptr, int dmin,
- int xmin, int ymin, int xmax, int ymax,
- int pred_x, int pred_y, Picture *ref_picture,
- int n, int size, uint8_t * const mv_penalty);
- int (*motion_search[7])(struct MpegEncContext * s, int block,
- int *mx_ptr, int *my_ptr,
- int P[10][2], int pred_x, int pred_y,
- int xmin, int ymin, int xmax, int ymax, Picture *ref_picture, int16_t (*last_mv)[2],
- int ref_mv_scale, uint8_t * const mv_penalty);
- int (*pre_motion_search)(struct MpegEncContext * s, int block,
- int *mx_ptr, int *my_ptr,
- int P[10][2], int pred_x, int pred_y,
- int xmin, int ymin, int xmax, int ymax, Picture *ref_picture, int16_t (*last_mv)[2],
- int ref_mv_scale, uint8_t * const mv_penalty);
- int (*get_mb_score)(struct MpegEncContext * s, int mx, int my, int pred_x, int pred_y, Picture *ref_picture,
- uint8_t * const mv_penalty);
+ int src_index, int ref_index,
+ int size, int h);
}MotionEstContext;
/**
int h263_msmpeg4; ///< generate MSMPEG4 compatible stream (deprecated, use msmpeg4_version instead)
int h263_flv; ///< use flv h263 header
- int codec_id; /* see CODEC_ID_xxx */
+ enum CodecID codec_id; /* see CODEC_ID_xxx */
int fixed_qscale; ///< fixed qscale if non zero
int encoding; ///< true if we are encoding (vs decoding)
int flags; ///< AVCodecContext.flags (HQ, MV4, ...)
+ int flags2; ///< AVCodecContext.flags2
int max_b_frames; ///< max number of b-frames for encoding
int luma_elim_threshold;
int chroma_elim_threshold;
int picture_number; //FIXME remove, unclear definition
int picture_in_gop_number; ///< 0-> first pic in gop, ...
int b_frames_since_non_b; ///< used for encoding, relative to not yet reordered input
+ int64_t user_specified_pts;///< last non zero pts from AVFrame which was passed into avcodec_encode_video()
int mb_width, mb_height; ///< number of MBs horizontally & vertically
int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressng of left & top MBs withoutt sig11
int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressng
Picture **input_picture; ///< next pictures on display order for encoding
Picture **reordered_input_picture; ///< pointer to the next pictures in codedorder for encoding
+ int start_mb_y; ///< start mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
+ int end_mb_y; ///< end mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
+ struct MpegEncContext *thread_context[MAX_THREADS];
+
/**
* copy of the previous picture structure.
* note, linesize & data, might not match the previous picture (for field pictures)
Picture *last_picture_ptr; ///< pointer to the previous picture.
Picture *next_picture_ptr; ///< pointer to the next picture (for bidir pred)
Picture *current_picture_ptr; ///< pointer to the current picture
+ uint8_t *visualization_buffer[3]; //< temporary buffer vor MV visualization
int last_dc[3]; ///< last DC values for MPEG1
+ int16_t *dc_val_base;
int16_t *dc_val[3]; ///< used for mpeg4 DC prediction, all 3 arrays must be continuous
int16_t dc_cache[4*5];
int y_dc_scale, c_dc_scale;
uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table
uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table
const uint8_t *chroma_qscale_table; ///< qscale -> chroma_qscale (h263)
+ uint8_t *coded_block_base;
uint8_t *coded_block; ///< used for coded block pattern prediction (msmpeg4v3, wmv1)
+ int16_t (*ac_val_base)[16];
int16_t (*ac_val[3])[16]; ///< used for for mpeg4 AC prediction, all 3 arrays must be continuous
int ac_pred;
uint8_t *prev_pict_types; ///< previous picture types in bitstream order, used for mb skip
uint8_t *cbp_table; ///< used to store cbp, ac_pred for partitioned decoding
uint8_t *pred_dir_table; ///< used to store pred_dir for partitioned decoding
uint8_t *allocated_edge_emu_buffer;
- uint8_t *edge_emu_buffer; ///< points into the middle of allocated_edge_emu_buffer
+ uint8_t *edge_emu_buffer; ///< points into the middle of allocated_edge_emu_buffer
+ uint8_t *rd_scratchpad; ///< scartchpad for rate distortion mb decission
+ uint8_t *obmc_scratchpad;
+ uint8_t *b_scratchpad; ///< scratchpad used for writing into write only buffers
int qscale; ///< QP
int chroma_qscale; ///< chroma QP
int adaptive_quant; ///< use adaptive quantization
int dquant; ///< qscale difference to prev qscale
int pict_type; ///< I_TYPE, P_TYPE, B_TYPE, ...
- int last_pict_type;
+ int last_pict_type; //FIXME removes
int last_non_b_pict_type; ///< used for mpeg4 gmc b-frames & ratecontrol
+ int dropable;
int frame_rate_index;
+
/* motion compensation */
int unrestricted_mv; ///< mv can point outside of the coded picture
int h263_long_vectors; ///< use horrible h263v1 long vector mode
int16_t (*b_bidir_forw_mv_table_base)[2];
int16_t (*b_bidir_back_mv_table_base)[2];
int16_t (*b_direct_mv_table_base)[2];
+ int16_t (*p_field_mv_table_base[2][2])[2];
+ int16_t (*b_field_mv_table_base[2][2][2])[2];
int16_t (*p_mv_table)[2]; ///< MV table (1MV per MB) p-frame encoding
int16_t (*b_forw_mv_table)[2]; ///< MV table (1MV per MB) forward mode b-frame encoding
int16_t (*b_back_mv_table)[2]; ///< MV table (1MV per MB) backward mode b-frame encoding
int16_t (*b_bidir_forw_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
int16_t (*b_bidir_back_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
int16_t (*b_direct_mv_table)[2]; ///< MV table (1MV per MB) direct mode b-frame encoding
+ int16_t (*p_field_mv_table[2][2])[2]; ///< MV table (2MV per MB) interlaced p-frame encoding
+ int16_t (*b_field_mv_table[2][2][2])[2];///< MV table (4MV per MB) interlaced b-frame encoding
+ uint8_t (*p_field_select_table[2]);
+ uint8_t (*b_field_select_table[2][2]);
int me_method; ///< ME algorithm
- int scene_change_score;
int mv_dir;
#define MV_DIR_BACKWARD 1
#define MV_DIR_FORWARD 2
int mb_x, mb_y;
int mb_skip_run;
int mb_intra;
- uint8_t *mb_type; ///< Table for MB type FIXME remove and use picture->mb_type
-#define MB_TYPE_INTRA 0x01
-#define MB_TYPE_INTER 0x02
-#define MB_TYPE_INTER4V 0x04
-#define MB_TYPE_SKIPED 0x08
+ uint16_t *mb_type; ///< Table for candidate MB types for encoding
+#define CANDIDATE_MB_TYPE_INTRA 0x01
+#define CANDIDATE_MB_TYPE_INTER 0x02
+#define CANDIDATE_MB_TYPE_INTER4V 0x04
+#define CANDIDATE_MB_TYPE_SKIPED 0x08
//#define MB_TYPE_GMC 0x10
-#define MB_TYPE_DIRECT 0x10
-#define MB_TYPE_FORWARD 0x20
-#define MB_TYPE_BACKWARD 0x40
-#define MB_TYPE_BIDIR 0x80
+#define CANDIDATE_MB_TYPE_DIRECT 0x10
+#define CANDIDATE_MB_TYPE_FORWARD 0x20
+#define CANDIDATE_MB_TYPE_BACKWARD 0x40
+#define CANDIDATE_MB_TYPE_BIDIR 0x80
+
+#define CANDIDATE_MB_TYPE_INTER_I 0x100
+#define CANDIDATE_MB_TYPE_FORWARD_I 0x200
+#define CANDIDATE_MB_TYPE_BACKWARD_I 0x400
+#define CANDIDATE_MB_TYPE_BIDIR_I 0x800
int block_index[6]; ///< index to current MB in block based arrays with edges
int block_wrap[6];
/** identical to the above but for MMX & these are not permutated, second 64 entries are bias*/
uint16_t (*q_intra_matrix16)[2][64];
uint16_t (*q_inter_matrix16)[2][64];
- int block_last_index[6]; ///< last non zero coefficient in block
+ int block_last_index[12]; ///< last non zero coefficient in block
/* scantables */
ScanTable __align8 intra_scantable;
ScanTable intra_h_scantable;
void *opaque; ///< private data for the user
/* bit rate control */
- int I_frame_bits; //FIXME used in mpeg12 ...
int64_t wanted_bits;
int64_t total_bits;
int frame_bits; ///< bits used for the current frame
int alt_inter_vlc; ///< alternative inter vlc
int modified_quant;
int loop_filter;
+ int custom_pcf;
/* mpeg4 specific */
int time_increment_resolution;
int intra_dc_threshold; ///< QP above whch the ac VLC should be used for intra dc
PutBitContext tex_pb; ///< used for data partitioned VOPs
PutBitContext pb2; ///< used for data partitioned VOPs
-#define PB_BUFFER_SIZE 1024*256
- uint8_t *tex_pb_buffer;
- uint8_t *pb2_buffer;
int mpeg_quant;
- int16_t (*field_mv_table)[2][2]; ///< used for interlaced b frame decoding
- int8_t (*field_select_table)[2]; ///< wtf, no really another table for interlaced b frames
int t_frame; ///< time distance of first I -> B, used for interlaced b frames
int padding_bug_score; ///< used to detect the VERY common padding bug in MPEG4
int divx_version;
int divx_build;
int divx_packed;
-#define BITSTREAM_BUFFER_SIZE 1024*256
uint8_t *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
int bitstream_buffer_size;
+ int allocated_bitstream_buffer_size;
int xvid_build;
GetBitContext gb;
/* Mpeg1 specific */
- int fake_picture_number; ///< picture number at the bitstream frame rate
int gop_picture_number; ///< index of the first picture of a GOP based on fake_pic_num & mpeg1 specific
int last_mv_dir; ///< last mv_dir, used for b frame encoding
int broken_link; ///< no_output_of_prior_pics_flag
int alternate_scan;
int repeat_first_field;
int chroma_420_type;
+ int chroma_format;
+#define CHROMA_420 1
+#define CHROMA_422 2
+#define CHROMA_444 3
+ int chroma_x_shift;//depend on pix_format, that depend on chroma_format
+ int chroma_y_shift;
+
int progressive_frame;
int full_pel[2];
int interlaced_dct;
DCTELEM *block/*align 16*/, int n, int qscale);
void (*dct_unquantize_h263_inter)(struct MpegEncContext *s,
DCTELEM *block/*align 16*/, int n, int qscale);
+ void (*dct_unquantize_h261_intra)(struct MpegEncContext *s,
+ DCTELEM *block/*align 16*/, int n, int qscale);
+ void (*dct_unquantize_h261_inter)(struct MpegEncContext *s,
+ DCTELEM *block/*align 16*/, int n, int qscale);
void (*dct_unquantize_intra)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
DCTELEM *block/*align 16*/, int n, int qscale);
void (*dct_unquantize_inter)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
DCTELEM *block/*align 16*/, int n, int qscale);
int (*dct_quantize)(struct MpegEncContext *s, DCTELEM *block/*align 16*/, int n, int qscale, int *overflow);
int (*fast_dct_quantize)(struct MpegEncContext *s, DCTELEM *block/*align 16*/, int n, int qscale, int *overflow);
+ void (*denoise_dct)(struct MpegEncContext *s, DCTELEM *block);
} MpegEncContext;
int DCT_common_init(MpegEncContext *s);
+void MPV_decode_defaults(MpegEncContext *s);
int MPV_common_init(MpegEncContext *s);
void MPV_common_end(MpegEncContext *s);
-void MPV_decode_mb(MpegEncContext *s, DCTELEM block[6][64]);
+void MPV_decode_mb(MpegEncContext *s, DCTELEM block[12][64]);
int MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx);
void MPV_frame_end(MpegEncContext *s);
int MPV_encode_init(AVCodecContext *avctx);
void ff_emulated_edge_mc(uint8_t *buf, uint8_t *src, int linesize, int block_w, int block_h,
int src_x, int src_y, int w, int h);
#define END_NOT_FOUND -100
-int ff_combine_frame( MpegEncContext *s, int next, uint8_t **buf, int *buf_size);
+int ff_combine_frame(ParseContext *pc, int next, uint8_t **buf, int *buf_size);
+void ff_parse_close(AVCodecParserContext *s);
void ff_mpeg_flush(AVCodecContext *avctx);
-void ff_print_debug_info(MpegEncContext *s, Picture *pict);
+void ff_print_debug_info(MpegEncContext *s, AVFrame *pict);
void ff_write_quant_matrix(PutBitContext *pb, int16_t *matrix);
int ff_find_unused_picture(MpegEncContext *s, int shared);
void ff_denoise_dct(MpegEncContext *s, DCTELEM *block);
+void ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
void ff_er_frame_start(MpegEncContext *s);
void ff_er_frame_end(MpegEncContext *s);
void ff_init_block_index(MpegEncContext *s);
static inline void ff_update_block_index(MpegEncContext *s){
+ const int block_size= 8>>s->avctx->lowres;
+
s->block_index[0]+=2;
s->block_index[1]+=2;
s->block_index[2]+=2;
s->block_index[3]+=2;
s->block_index[4]++;
s->block_index[5]++;
- s->dest[0]+= 16;
- s->dest[1]+= 8;
- s->dest[2]+= 8;
+ s->dest[0]+= 2*block_size;
+ s->dest[1]+= block_size;
+ s->dest[2]+= block_size;
}
static inline int get_bits_diff(MpegEncContext *s){
- const int bits= get_bit_count(&s->pb);
+ const int bits= put_bits_count(&s->pb);
const int last= s->last_bits;
s->last_bits = bits;
int mb_x, int mb_y);
int ff_get_best_fcode(MpegEncContext * s, int16_t (*mv_table)[2], int type);
void ff_fix_long_p_mvs(MpegEncContext * s);
-void ff_fix_long_b_mvs(MpegEncContext * s, int16_t (*mv_table)[2], int f_code, int type);
+void ff_fix_long_mvs(MpegEncContext * s, uint8_t *field_select_table, int field_select,
+ int16_t (*mv_table)[2], int f_code, int type, int truncate);
void ff_init_me(MpegEncContext *s);
int ff_pre_estimate_p_frame_motion(MpegEncContext * s, int mb_x, int mb_y);
-
+inline int ff_epzs_motion_search(MpegEncContext * s, int *mx_ptr, int *my_ptr,
+ int P[10][2], int src_index, int ref_index, int16_t (*last_mv)[2],
+ int ref_mv_scale, int size, int h);
+int inline ff_get_mb_score(MpegEncContext * s, int mx, int my, int src_index,
+ int ref_index, int size, int h, int add_rate);
/* mpeg12.c */
extern const int16_t ff_mpeg1_default_intra_matrix[64];
void ff_mpeg1_encode_init(MpegEncContext *s);
void ff_mpeg1_encode_slice_header(MpegEncContext *s);
void ff_mpeg1_clean_buffers(MpegEncContext *s);
+int ff_mpeg1_find_frame_end(ParseContext *pc, const uint8_t *buf, int buf_size);
/** RLTable. */
RL_VLC_ELEM *rl_vlc[32]; ///< decoding only
} RLTable;
-void init_rl(RLTable *rl);
-void init_vlc_rl(RLTable *rl);
+void init_rl(RLTable *rl, int use_static);
+void init_vlc_rl(RLTable *rl, int use_static);
static inline int get_rl_index(const RLTable *rl, int last, int run, int level)
{
extern const uint8_t ff_h263_chroma_qscale_table[32];
extern const uint8_t ff_h263_loop_filter_strength[32];
+/* h261.c */
+void ff_h261_loop_filter(MpegEncContext *s);
+void ff_h261_reorder_mb_index(MpegEncContext* s);
+void ff_h261_encode_mb(MpegEncContext *s,
+ DCTELEM block[6][64],
+ int motion_x, int motion_y);
+void ff_h261_encode_picture_header(MpegEncContext * s, int picture_number);
+void ff_h261_encode_init(MpegEncContext *s);
+
+/* h263.c, h263dec.c */
int ff_h263_decode_init(AVCodecContext *avctx);
int ff_h263_decode_frame(AVCodecContext *avctx,
void *data, int *data_size,
void h263_encode_picture_header(MpegEncContext *s, int picture_number);
void ff_flv_encode_picture_header(MpegEncContext *s, int picture_number);
void h263_encode_gob_header(MpegEncContext * s, int mb_line);
-int16_t *h263_pred_motion(MpegEncContext * s, int block,
+int16_t *h263_pred_motion(MpegEncContext * s, int block, int dir,
int *px, int *py);
void mpeg4_pred_ac(MpegEncContext * s, DCTELEM *block, int n,
int dir);
int ff_h263_resync(MpegEncContext *s);
int ff_h263_get_gob_height(MpegEncContext *s);
int ff_mpeg4_set_direct_mv(MpegEncContext *s, int mx, int my);
-inline int ff_h263_round_chroma(int x);
+int ff_h263_round_chroma(int x);
+void ff_h263_encode_motion(MpegEncContext * s, int val, int f_code);
+int ff_mpeg4_find_frame_end(ParseContext *pc, const uint8_t *buf, int buf_size);
/* rv10.c */
void rv10_encode_picture_header(MpegEncContext *s, int picture_number);
int rv_decode_dc(MpegEncContext *s, int n);
+void rv20_encode_picture_header(MpegEncContext *s, int picture_number);
/* msmpeg4.c */
DCTELEM block[6][64]);
void mjpeg_picture_header(MpegEncContext *s);
void mjpeg_picture_trailer(MpegEncContext *s);
+void ff_mjpeg_stuffing(PutBitContext * pbc);
/* rate control */
projects / ffmpeg / blobdiff