yading@10: /*
yading@10:  * Zip Motion Blocks Video (ZMBV) encoder
yading@10:  * Copyright (c) 2006 Konstantin Shishkov
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: /**
yading@10:  * @file
yading@10:  * Zip Motion Blocks Video encoder
yading@10:  */
yading@10: 
yading@10: #include <stdio.h>
yading@10: #include <stdlib.h>
yading@10: 
yading@10: #include "libavutil/common.h"
yading@10: #include "libavutil/intreadwrite.h"
yading@10: #include "avcodec.h"
yading@10: #include "internal.h"
yading@10: 
yading@10: #include <zlib.h>
yading@10: 
yading@10: #define ZMBV_KEYFRAME 1
yading@10: #define ZMBV_DELTAPAL 2
yading@10: 
yading@10: #define ZMBV_BLOCK 16
yading@10: 
yading@10: /**
yading@10:  * Encoder context
yading@10:  */
yading@10: typedef struct ZmbvEncContext {
yading@10:     AVCodecContext *avctx;
yading@10:     AVFrame pic;
yading@10: 
yading@10:     int range;
yading@10:     uint8_t *comp_buf, *work_buf;
yading@10:     uint8_t pal[768];
yading@10:     uint32_t pal2[256]; //for quick comparisons
yading@10:     uint8_t *prev;
yading@10:     int pstride;
yading@10:     int comp_size;
yading@10:     int keyint, curfrm;
yading@10:     z_stream zstream;
yading@10: } ZmbvEncContext;
yading@10: 
yading@10: static int score_tab[256];
yading@10: 
yading@10: /** Block comparing function
yading@10:  * XXX should be optimized and moved to DSPContext
yading@10:  * TODO handle out of edge ME
yading@10:  */
yading@10: static inline int block_cmp(uint8_t *src, int stride, uint8_t *src2, int stride2,
yading@10:                             int bw, int bh, int *xored)
yading@10: {
yading@10:     int sum = 0;
yading@10:     int i, j;
yading@10:     uint8_t histogram[256] = {0};
yading@10: 
yading@10:     *xored = 0;
yading@10:     for(j = 0; j < bh; j++){
yading@10:         for(i = 0; i < bw; i++){
yading@10:             int t = src[i] ^ src2[i];
yading@10:             histogram[t]++;
yading@10:             *xored |= t;
yading@10:         }
yading@10:         src += stride;
yading@10:         src2 += stride2;
yading@10:     }
yading@10: 
yading@10:     for(i = 1; i < 256; i++)
yading@10:         sum += score_tab[histogram[i]];
yading@10: 
yading@10:     return sum;
yading@10: }
yading@10: 
yading@10: /** Motion estimation function
yading@10:  * TODO make better ME decisions
yading@10:  */
yading@10: static int zmbv_me(ZmbvEncContext *c, uint8_t *src, int sstride, uint8_t *prev,
yading@10:                    int pstride, int x, int y, int *mx, int *my, int *xored)
yading@10: {
yading@10:     int dx, dy, tx, ty, tv, bv, bw, bh;
yading@10: 
yading@10:     *mx = *my = 0;
yading@10:     bw = FFMIN(ZMBV_BLOCK, c->avctx->width - x);
yading@10:     bh = FFMIN(ZMBV_BLOCK, c->avctx->height - y);
yading@10:     bv = block_cmp(src, sstride, prev, pstride, bw, bh, xored);
yading@10:     if(!bv) return 0;
yading@10:     for(ty = FFMAX(y - c->range, 0); ty < FFMIN(y + c->range, c->avctx->height - bh); ty++){
yading@10:         for(tx = FFMAX(x - c->range, 0); tx < FFMIN(x + c->range, c->avctx->width - bw); tx++){
yading@10:             if(tx == x && ty == y) continue; // we already tested this block
yading@10:             dx = tx - x;
yading@10:             dy = ty - y;
yading@10:             tv = block_cmp(src, sstride, prev + dx + dy*pstride, pstride, bw, bh, xored);
yading@10:             if(tv < bv){
yading@10:                  bv = tv;
yading@10:                  *mx = dx;
yading@10:                  *my = dy;
yading@10:                  if(!bv) return 0;
yading@10:              }
yading@10:          }
yading@10:     }
yading@10:     return bv;
yading@10: }
yading@10: 
yading@10: static int encode_frame(AVCodecContext *avctx, AVPacket *pkt,
yading@10:                         const AVFrame *pict, int *got_packet)
yading@10: {
yading@10:     ZmbvEncContext * const c = avctx->priv_data;
yading@10:     AVFrame * const p = &c->pic;
yading@10:     uint8_t *src, *prev, *buf;
yading@10:     uint32_t *palptr;
yading@10:     int keyframe, chpal;
yading@10:     int fl;
yading@10:     int work_size = 0, pkt_size;
yading@10:     int bw, bh;
yading@10:     int i, j, ret;
yading@10: 
yading@10:     keyframe = !c->curfrm;
yading@10:     c->curfrm++;
yading@10:     if(c->curfrm == c->keyint)
yading@10:         c->curfrm = 0;
yading@10:     *p = *pict;
yading@10:     p->pict_type= keyframe ? AV_PICTURE_TYPE_I : AV_PICTURE_TYPE_P;
yading@10:     p->key_frame= keyframe;
yading@10:     chpal = !keyframe && memcmp(p->data[1], c->pal2, 1024);
yading@10: 
yading@10:     palptr = (uint32_t*)p->data[1];
yading@10:     src = p->data[0];
yading@10:     prev = c->prev;
yading@10:     if(chpal){
yading@10:         uint8_t tpal[3];
yading@10:         for(i = 0; i < 256; i++){
yading@10:             AV_WB24(tpal, palptr[i]);
yading@10:             c->work_buf[work_size++] = tpal[0] ^ c->pal[i * 3 + 0];
yading@10:             c->work_buf[work_size++] = tpal[1] ^ c->pal[i * 3 + 1];
yading@10:             c->work_buf[work_size++] = tpal[2] ^ c->pal[i * 3 + 2];
yading@10:             c->pal[i * 3 + 0] = tpal[0];
yading@10:             c->pal[i * 3 + 1] = tpal[1];
yading@10:             c->pal[i * 3 + 2] = tpal[2];
yading@10:         }
yading@10:         memcpy(c->pal2, p->data[1], 1024);
yading@10:     }
yading@10:     if(keyframe){
yading@10:         for(i = 0; i < 256; i++){
yading@10:             AV_WB24(c->pal+(i*3), palptr[i]);
yading@10:         }
yading@10:         memcpy(c->work_buf, c->pal, 768);
yading@10:         memcpy(c->pal2, p->data[1], 1024);
yading@10:         work_size = 768;
yading@10:         for(i = 0; i < avctx->height; i++){
yading@10:             memcpy(c->work_buf + work_size, src, avctx->width);
yading@10:             src += p->linesize[0];
yading@10:             work_size += avctx->width;
yading@10:         }
yading@10:     }else{
yading@10:         int x, y, bh2, bw2, xored;
yading@10:         uint8_t *tsrc, *tprev;
yading@10:         uint8_t *mv;
yading@10:         int mx, my;
yading@10: 
yading@10:         bw = (avctx->width + ZMBV_BLOCK - 1) / ZMBV_BLOCK;
yading@10:         bh = (avctx->height + ZMBV_BLOCK - 1) / ZMBV_BLOCK;
yading@10:         mv = c->work_buf + work_size;
yading@10:         memset(c->work_buf + work_size, 0, (bw * bh * 2 + 3) & ~3);
yading@10:         work_size += (bw * bh * 2 + 3) & ~3;
yading@10:         /* for now just XOR'ing */
yading@10:         for(y = 0; y < avctx->height; y += ZMBV_BLOCK) {
yading@10:             bh2 = FFMIN(avctx->height - y, ZMBV_BLOCK);
yading@10:             for(x = 0; x < avctx->width; x += ZMBV_BLOCK, mv += 2) {
yading@10:                 bw2 = FFMIN(avctx->width - x, ZMBV_BLOCK);
yading@10: 
yading@10:                 tsrc = src + x;
yading@10:                 tprev = prev + x;
yading@10: 
yading@10:                 zmbv_me(c, tsrc, p->linesize[0], tprev, c->pstride, x, y, &mx, &my, &xored);
yading@10:                 mv[0] = (mx << 1) | !!xored;
yading@10:                 mv[1] = my << 1;
yading@10:                 tprev += mx + my * c->pstride;
yading@10:                 if(xored){
yading@10:                     for(j = 0; j < bh2; j++){
yading@10:                         for(i = 0; i < bw2; i++)
yading@10:                             c->work_buf[work_size++] = tsrc[i] ^ tprev[i];
yading@10:                         tsrc += p->linesize[0];
yading@10:                         tprev += c->pstride;
yading@10:                     }
yading@10:                 }
yading@10:             }
yading@10:             src += p->linesize[0] * ZMBV_BLOCK;
yading@10:             prev += c->pstride * ZMBV_BLOCK;
yading@10:         }
yading@10:     }
yading@10:     /* save the previous frame */
yading@10:     src = p->data[0];
yading@10:     prev = c->prev;
yading@10:     for(i = 0; i < avctx->height; i++){
yading@10:         memcpy(prev, src, avctx->width);
yading@10:         prev += c->pstride;
yading@10:         src += p->linesize[0];
yading@10:     }
yading@10: 
yading@10:     if (keyframe)
yading@10:         deflateReset(&c->zstream);
yading@10: 
yading@10:     c->zstream.next_in = c->work_buf;
yading@10:     c->zstream.avail_in = work_size;
yading@10:     c->zstream.total_in = 0;
yading@10: 
yading@10:     c->zstream.next_out = c->comp_buf;
yading@10:     c->zstream.avail_out = c->comp_size;
yading@10:     c->zstream.total_out = 0;
yading@10:     if(deflate(&c->zstream, Z_SYNC_FLUSH) != Z_OK){
yading@10:         av_log(avctx, AV_LOG_ERROR, "Error compressing data\n");
yading@10:         return -1;
yading@10:     }
yading@10: 
yading@10:     pkt_size = c->zstream.total_out + 1 + 6*keyframe;
yading@10:     if ((ret = ff_alloc_packet2(avctx, pkt, pkt_size)) < 0)
yading@10:         return ret;
yading@10:     buf = pkt->data;
yading@10: 
yading@10:     fl = (keyframe ? ZMBV_KEYFRAME : 0) | (chpal ? ZMBV_DELTAPAL : 0);
yading@10:     *buf++ = fl;
yading@10:     if (keyframe) {
yading@10:         *buf++ = 0; // hi ver
yading@10:         *buf++ = 1; // lo ver
yading@10:         *buf++ = 1; // comp
yading@10:         *buf++ = 4; // format - 8bpp
yading@10:         *buf++ = ZMBV_BLOCK; // block width
yading@10:         *buf++ = ZMBV_BLOCK; // block height
yading@10:     }
yading@10:     memcpy(buf, c->comp_buf, c->zstream.total_out);
yading@10: 
yading@10:     pkt->flags |= AV_PKT_FLAG_KEY*keyframe;
yading@10:     *got_packet = 1;
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: 
yading@10: /**
yading@10:  * Init zmbv encoder
yading@10:  */
yading@10: static av_cold int encode_init(AVCodecContext *avctx)
yading@10: {
yading@10:     ZmbvEncContext * const c = avctx->priv_data;
yading@10:     int zret; // Zlib return code
yading@10:     int i;
yading@10:     int lvl = 9;
yading@10: 
yading@10:     for(i=1; i<256; i++)
yading@10:         score_tab[i]= -i * log(i/(double)(ZMBV_BLOCK*ZMBV_BLOCK)) * (256/M_LN2);
yading@10: 
yading@10:     c->avctx = avctx;
yading@10: 
yading@10:     c->curfrm = 0;
yading@10:     c->keyint = avctx->keyint_min;
yading@10:     c->range = 8;
yading@10:     if(avctx->me_range > 0)
yading@10:         c->range = FFMIN(avctx->me_range, 127);
yading@10: 
yading@10:     if(avctx->compression_level >= 0)
yading@10:         lvl = avctx->compression_level;
yading@10:     if(lvl < 0 || lvl > 9){
yading@10:         av_log(avctx, AV_LOG_ERROR, "Compression level should be 0-9, not %i\n", lvl);
yading@10:         return AVERROR(EINVAL);
yading@10:     }
yading@10: 
yading@10:     // Needed if zlib unused or init aborted before deflateInit
yading@10:     memset(&c->zstream, 0, sizeof(z_stream));
yading@10:     c->comp_size = avctx->width * avctx->height + 1024 +
yading@10:         ((avctx->width + ZMBV_BLOCK - 1) / ZMBV_BLOCK) * ((avctx->height + ZMBV_BLOCK - 1) / ZMBV_BLOCK) * 2 + 4;
yading@10:     if ((c->work_buf = av_malloc(c->comp_size)) == NULL) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "Can't allocate work buffer.\n");
yading@10:         return AVERROR(ENOMEM);
yading@10:     }
yading@10:     /* Conservative upper bound taken from zlib v1.2.1 source via lcl.c */
yading@10:     c->comp_size = c->comp_size + ((c->comp_size + 7) >> 3) +
yading@10:                            ((c->comp_size + 63) >> 6) + 11;
yading@10: 
yading@10:     /* Allocate compression buffer */
yading@10:     if ((c->comp_buf = av_malloc(c->comp_size)) == NULL) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "Can't allocate compression buffer.\n");
yading@10:         return AVERROR(ENOMEM);
yading@10:     }
yading@10:     c->pstride = FFALIGN(avctx->width, 16);
yading@10:     if ((c->prev = av_malloc(c->pstride * avctx->height)) == NULL) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "Can't allocate picture.\n");
yading@10:         return AVERROR(ENOMEM);
yading@10:     }
yading@10: 
yading@10:     c->zstream.zalloc = Z_NULL;
yading@10:     c->zstream.zfree = Z_NULL;
yading@10:     c->zstream.opaque = Z_NULL;
yading@10:     zret = deflateInit(&c->zstream, lvl);
yading@10:     if (zret != Z_OK) {
yading@10:         av_log(avctx, AV_LOG_ERROR, "Inflate init error: %d\n", zret);
yading@10:         return -1;
yading@10:     }
yading@10: 
yading@10:     avctx->coded_frame = &c->pic;
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: 
yading@10: 
yading@10: /**
yading@10:  * Uninit zmbv encoder
yading@10:  */
yading@10: static av_cold int encode_end(AVCodecContext *avctx)
yading@10: {
yading@10:     ZmbvEncContext * const c = avctx->priv_data;
yading@10: 
yading@10:     av_freep(&c->comp_buf);
yading@10:     av_freep(&c->work_buf);
yading@10: 
yading@10:     deflateEnd(&c->zstream);
yading@10:     av_freep(&c->prev);
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: AVCodec ff_zmbv_encoder = {
yading@10:     .name           = "zmbv",
yading@10:     .type           = AVMEDIA_TYPE_VIDEO,
yading@10:     .id             = AV_CODEC_ID_ZMBV,
yading@10:     .priv_data_size = sizeof(ZmbvEncContext),
yading@10:     .init           = encode_init,
yading@10:     .encode2        = encode_frame,
yading@10:     .close          = encode_end,
yading@10:     .pix_fmts       = (const enum AVPixelFormat[]){ AV_PIX_FMT_PAL8, AV_PIX_FMT_NONE },
yading@10:     .long_name      = NULL_IF_CONFIG_SMALL("Zip Motion Blocks Video"),
yading@10: };