view flashsvenc.c @ 11285:613370892df2 libavcodec

Store intra4x4_pred_mode per row only. about 5 cpu cycles slower in the local code but should be overall faster due to reduced cache use. (my sample though has too few intra4x4 blocks for this to be meassureable easily either way)

/*
 * Flash Screen Video encoder
 * Copyright (C) 2004 Alex Beregszaszi
 * Copyright (C) 2006 Benjamin Larsson
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/* Encoding development sponsored by http://fh-campuswien.ac.at */

/**
 * @file libavcodec/flashsvenc.c
 * Flash Screen Video encoder
 * @author Alex Beregszaszi
 * @author Benjamin Larsson
 */

/* Bitstream description
 * The picture is divided into blocks that are zlib-compressed.
 *
 * The decoder is fed complete frames, the frameheader contains:
 * 4bits of block width
 * 12bits of frame width
 * 4bits of block height
 * 12bits of frame height
 *
 * Directly after the header are the compressed blocks. The blocks
 * have their compressed size represented with 16bits in the beginig.
 * If the size = 0 then the block is unchanged from the previous frame.
 * All blocks are decompressed until the buffer is consumed.
 *
 * Encoding ideas, a basic encoder would just use a fixed block size.
 * Block sizes can be multipels of 16, from 16 to 256. The blocks don't
 * have to be quadratic. A brute force search with a set of different
 * block sizes should give a better result than to just use a fixed size.
 */

/* TODO:
 * Don't reencode the frame in brute force mode if the frame is a dupe. Speed up.
 * Make the difference check faster.
 */

#include <stdio.h>
#include <stdlib.h>
#include <zlib.h>

#include "avcodec.h"
#include "put_bits.h"
#include "bytestream.h"


typedef struct FlashSVContext {
    AVCodecContext *avctx;
    uint8_t *previous_frame;
    AVFrame frame;
    int image_width, image_height;
    int block_width, block_height;
    uint8_t* tmpblock;
    uint8_t* encbuffer;
    int block_size;
    z_stream zstream;
    int last_key_frame;
} FlashSVContext;

static int copy_region_enc(uint8_t *sptr, uint8_t *dptr,
        int dx, int dy, int h, int w, int stride, uint8_t *pfptr) {
    int i,j;
    uint8_t *nsptr;
    uint8_t *npfptr;
    int diff = 0;

    for (i = dx+h; i > dx; i--) {
        nsptr = sptr+(i*stride)+dy*3;
        npfptr = pfptr+(i*stride)+dy*3;
        for (j=0 ; j<w*3 ; j++) {
            diff |=npfptr[j]^nsptr[j];
            dptr[j] = nsptr[j];
        }
        dptr += w*3;
    }
    if (diff)
        return 1;
    return 0;
}

static av_cold int flashsv_encode_init(AVCodecContext *avctx)
{
    FlashSVContext *s = avctx->priv_data;

    s->avctx = avctx;

    if ((avctx->width > 4095) || (avctx->height > 4095)) {
        av_log(avctx, AV_LOG_ERROR, "Input dimensions too large, input must be max 4096x4096 !\n");
        return -1;
    }

    // Needed if zlib unused or init aborted before deflateInit
    memset(&(s->zstream), 0, sizeof(z_stream));

    s->last_key_frame=0;

    s->image_width = avctx->width;
    s->image_height = avctx->height;

    s->tmpblock = av_mallocz(3*256*256);
    s->encbuffer = av_mallocz(s->image_width*s->image_height*3);

    if (!s->tmpblock || !s->encbuffer) {
        av_log(avctx, AV_LOG_ERROR, "Memory allocation failed.\n");
        return -1;
    }

    return 0;
}


static int encode_bitstream(FlashSVContext *s, AVFrame *p, uint8_t *buf, int buf_size,
     int block_width, int block_height, uint8_t *previous_frame, int* I_frame) {

    PutBitContext pb;
    int h_blocks, v_blocks, h_part, v_part, i, j;
    int buf_pos, res;
    int pred_blocks = 0;

    init_put_bits(&pb, buf, buf_size*8);

    put_bits(&pb, 4, (block_width/16)-1);
    put_bits(&pb, 12, s->image_width);
    put_bits(&pb, 4, (block_height/16)-1);
    put_bits(&pb, 12, s->image_height);
    flush_put_bits(&pb);
    buf_pos=4;

    h_blocks = s->image_width / block_width;
    h_part = s->image_width % block_width;
    v_blocks = s->image_height / block_height;
    v_part = s->image_height % block_height;

    /* loop over all block columns */
    for (j = 0; j < v_blocks + (v_part?1:0); j++)
    {

        int hp = j*block_height; // horiz position in frame
        int hs = (j<v_blocks)?block_height:v_part; // size of block

        /* loop over all block rows */
        for (i = 0; i < h_blocks + (h_part?1:0); i++)
        {
            int wp = i*block_width; // vert position in frame
            int ws = (i<h_blocks)?block_width:h_part; // size of block
            int ret=Z_OK;
            uint8_t *ptr;

            ptr = buf+buf_pos;

            //copy the block to the temp buffer before compression (if it differs from the previous frame's block)
            res = copy_region_enc(p->data[0], s->tmpblock, s->image_height-(hp+hs+1), wp, hs, ws, p->linesize[0], previous_frame);

            if (res || *I_frame) {
                unsigned long zsize;
                zsize = 3*block_width*block_height;
                ret = compress2(ptr+2, &zsize, s->tmpblock, 3*ws*hs, 9);


                //ret = deflateReset(&(s->zstream));
                if (ret != Z_OK)
                    av_log(s->avctx, AV_LOG_ERROR, "error while compressing block %dx%d\n", i, j);

                bytestream_put_be16(&ptr,(unsigned int)zsize);
                buf_pos += zsize+2;
                //av_log(avctx, AV_LOG_ERROR, "buf_pos = %d\n", buf_pos);
            } else {
                pred_blocks++;
                bytestream_put_be16(&ptr,0);
                buf_pos += 2;
            }
        }
    }

    if (pred_blocks)
        *I_frame = 0;
    else
        *I_frame = 1;

    return buf_pos;
}


static int flashsv_encode_frame(AVCodecContext *avctx, uint8_t *buf, int buf_size, void *data)
{
    FlashSVContext * const s = avctx->priv_data;
    AVFrame *pict = data;
    AVFrame * const p = &s->frame;
    uint8_t *pfptr;
    int res;
    int I_frame = 0;
    int opt_w, opt_h;

    *p = *pict;

    /* First frame needs to be a keyframe */
    if (avctx->frame_number == 0) {
        s->previous_frame = av_mallocz(FFABS(p->linesize[0])*s->image_height);
        if (!s->previous_frame) {
            av_log(avctx, AV_LOG_ERROR, "Memory allocation failed.\n");
            return -1;
        }
        I_frame = 1;
    }

    if (p->linesize[0] < 0)
        pfptr = s->previous_frame - ((s->image_height-1) * p->linesize[0]);
    else
        pfptr = s->previous_frame;

    /* Check the placement of keyframes */
    if (avctx->gop_size > 0) {
        if (avctx->frame_number >= s->last_key_frame + avctx->gop_size) {
            I_frame = 1;
        }
    }

    opt_w=4;
    opt_h=4;

    if (buf_size < s->image_width*s->image_height*3) {
        //Conservative upper bound check for compressed data
        av_log(avctx, AV_LOG_ERROR, "buf_size %d <  %d\n", buf_size, s->image_width*s->image_height*3);
        return -1;
    }

    res = encode_bitstream(s, p, buf, buf_size, opt_w*16, opt_h*16, pfptr, &I_frame);

    //save the current frame
    if(p->linesize[0] > 0)
        memcpy(s->previous_frame, p->data[0], s->image_height*p->linesize[0]);
    else
        memcpy(s->previous_frame, p->data[0] + p->linesize[0] * (s->image_height-1), s->image_height*FFABS(p->linesize[0]));

    //mark the frame type so the muxer can mux it correctly
    if (I_frame) {
        p->pict_type = FF_I_TYPE;
        p->key_frame = 1;
        s->last_key_frame = avctx->frame_number;
        av_log(avctx, AV_LOG_DEBUG, "Inserting key frame at frame %d\n",avctx->frame_number);
    } else {
        p->pict_type = FF_P_TYPE;
        p->key_frame = 0;
    }

    avctx->coded_frame = p;

    return res;
}

static av_cold int flashsv_encode_end(AVCodecContext *avctx)
{
    FlashSVContext *s = avctx->priv_data;

    deflateEnd(&(s->zstream));

    av_free(s->encbuffer);
    av_free(s->previous_frame);
    av_free(s->tmpblock);

    return 0;
}

AVCodec flashsv_encoder = {
    "flashsv",
    CODEC_TYPE_VIDEO,
    CODEC_ID_FLASHSV,
    sizeof(FlashSVContext),
    flashsv_encode_init,
    flashsv_encode_frame,
    flashsv_encode_end,
    .pix_fmts = (const enum PixelFormat[]){PIX_FMT_BGR24, PIX_FMT_NONE},
    .long_name = NULL_IF_CONFIG_SMALL("Flash Screen Video"),
};