FFmpeg: libavcodec/pngenc.c Source File

00001 /*
00002  * PNG image format
00003  * Copyright (c) 2003 Fabrice Bellard
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 #include "avcodec.h"
00022 #include "internal.h"
00023 #include "bytestream.h"
00024 #include "dsputil.h"
00025 #include "png.h"
00026 
00027 #include "libavutil/avassert.h"
00028 
00029 /* TODO:
00030  * - add 2, 4 and 16 bit depth support
00031  */
00032 
00033 #include <zlib.h>
00034 
00035 //#define DEBUG
00036 
00037 #define IOBUF_SIZE 4096
00038 
00039 typedef struct PNGEncContext {
00040     DSPContext dsp;
00041 
00042     uint8_t *bytestream;
00043     uint8_t *bytestream_start;
00044     uint8_t *bytestream_end;
00045     AVFrame picture;
00046 
00047     int filter_type;
00048 
00049     z_stream zstream;
00050     uint8_t buf[IOBUF_SIZE];
00051 } PNGEncContext;
00052 
00053 static void png_get_interlaced_row(uint8_t *dst, int row_size,
00054                                    int bits_per_pixel, int pass,
00055                                    const uint8_t *src, int width)
00056 {
00057     int x, mask, dst_x, j, b, bpp;
00058     uint8_t *d;
00059     const uint8_t *s;
00060 
00061     mask =  (int[]){0x80, 0x08, 0x88, 0x22, 0xaa, 0x55, 0xff}[pass];
00062     switch(bits_per_pixel) {
00063     case 1:
00064         memset(dst, 0, row_size);
00065         dst_x = 0;
00066         for(x = 0; x < width; x++) {
00067             j = (x & 7);
00068             if ((mask << j) & 0x80) {
00069                 b = (src[x >> 3] >> (7 - j)) & 1;
00070                 dst[dst_x >> 3] |= b << (7 - (dst_x & 7));
00071                 dst_x++;
00072             }
00073         }
00074         break;
00075     default:
00076         bpp = bits_per_pixel >> 3;
00077         d = dst;
00078         s = src;
00079         for(x = 0; x < width; x++) {
00080             j = x & 7;
00081             if ((mask << j) & 0x80) {
00082                 memcpy(d, s, bpp);
00083                 d += bpp;
00084             }
00085             s += bpp;
00086         }
00087         break;
00088     }
00089 }
00090 
00091 static void sub_png_paeth_prediction(uint8_t *dst, uint8_t *src, uint8_t *top, int w, int bpp)
00092 {
00093     int i;
00094     for(i = 0; i < w; i++) {
00095         int a, b, c, p, pa, pb, pc;
00096 
00097         a = src[i - bpp];
00098         b = top[i];
00099         c = top[i - bpp];
00100 
00101         p = b - c;
00102         pc = a - c;
00103 
00104         pa = abs(p);
00105         pb = abs(pc);
00106         pc = abs(p + pc);
00107 
00108         if (pa <= pb && pa <= pc)
00109             p = a;
00110         else if (pb <= pc)
00111             p = b;
00112         else
00113             p = c;
00114         dst[i] = src[i] - p;
00115     }
00116 }
00117 
00118 static void png_filter_row(DSPContext *dsp, uint8_t *dst, int filter_type,
00119                            uint8_t *src, uint8_t *top, int size, int bpp)
00120 {
00121     int i;
00122 
00123     switch(filter_type) {
00124     case PNG_FILTER_VALUE_NONE:
00125         memcpy(dst, src, size);
00126         break;
00127     case PNG_FILTER_VALUE_SUB:
00128         dsp->diff_bytes(dst, src, src-bpp, size);
00129         memcpy(dst, src, bpp);
00130         break;
00131     case PNG_FILTER_VALUE_UP:
00132         dsp->diff_bytes(dst, src, top, size);
00133         break;
00134     case PNG_FILTER_VALUE_AVG:
00135         for(i = 0; i < bpp; i++)
00136             dst[i] = src[i] - (top[i] >> 1);
00137         for(; i < size; i++)
00138             dst[i] = src[i] - ((src[i-bpp] + top[i]) >> 1);
00139         break;
00140     case PNG_FILTER_VALUE_PAETH:
00141         for(i = 0; i < bpp; i++)
00142             dst[i] = src[i] - top[i];
00143         sub_png_paeth_prediction(dst+i, src+i, top+i, size-i, bpp);
00144         break;
00145     }
00146 }
00147 
00148 static uint8_t *png_choose_filter(PNGEncContext *s, uint8_t *dst,
00149                                   uint8_t *src, uint8_t *top, int size, int bpp)
00150 {
00151     int pred = s->filter_type;
00152     av_assert0(bpp || !pred);
00153     if(!top && pred)
00154         pred = PNG_FILTER_VALUE_SUB;
00155     if(pred == PNG_FILTER_VALUE_MIXED) {
00156         int i;
00157         int cost, bcost = INT_MAX;
00158         uint8_t *buf1 = dst, *buf2 = dst + size + 16;
00159         for(pred=0; pred<5; pred++) {
00160             png_filter_row(&s->dsp, buf1+1, pred, src, top, size, bpp);
00161             buf1[0] = pred;
00162             cost = 0;
00163             for(i=0; i<=size; i++)
00164                 cost += abs((int8_t)buf1[i]);
00165             if(cost < bcost) {
00166                 bcost = cost;
00167                 FFSWAP(uint8_t*, buf1, buf2);
00168             }
00169         }
00170         return buf2;
00171     } else {
00172         png_filter_row(&s->dsp, dst+1, pred, src, top, size, bpp);
00173         dst[0] = pred;
00174         return dst;
00175     }
00176 }
00177 
00178 static void png_write_chunk(uint8_t **f, uint32_t tag,
00179                             const uint8_t *buf, int length)
00180 {
00181     uint32_t crc;
00182     uint8_t tagbuf[4];
00183 
00184     bytestream_put_be32(f, length);
00185     crc = crc32(0, Z_NULL, 0);
00186     AV_WL32(tagbuf, tag);
00187     crc = crc32(crc, tagbuf, 4);
00188     bytestream_put_be32(f, av_bswap32(tag));
00189     if (length > 0) {
00190         crc = crc32(crc, buf, length);
00191         memcpy(*f, buf, length);
00192         *f += length;
00193     }
00194     bytestream_put_be32(f, crc);
00195 }
00196 
00197 /* XXX: do filtering */
00198 static int png_write_row(PNGEncContext *s, const uint8_t *data, int size)
00199 {
00200     int ret;
00201 
00202     s->zstream.avail_in = size;
00203     s->zstream.next_in = (uint8_t *)data;
00204     while (s->zstream.avail_in > 0) {
00205         ret = deflate(&s->zstream, Z_NO_FLUSH);
00206         if (ret != Z_OK)
00207             return -1;
00208         if (s->zstream.avail_out == 0) {
00209             if(s->bytestream_end - s->bytestream > IOBUF_SIZE + 100)
00210                 png_write_chunk(&s->bytestream, MKTAG('I', 'D', 'A', 'T'), s->buf, IOBUF_SIZE);
00211             s->zstream.avail_out = IOBUF_SIZE;
00212             s->zstream.next_out = s->buf;
00213         }
00214     }
00215     return 0;
00216 }
00217 
00218 static int encode_frame(AVCodecContext *avctx, AVPacket *pkt,
00219                         const AVFrame *pict, int *got_packet)
00220 {
00221     PNGEncContext *s = avctx->priv_data;
00222     AVFrame * const p= &s->picture;
00223     int bit_depth, color_type, y, len, row_size, ret, is_progressive;
00224     int bits_per_pixel, pass_row_size, enc_row_size;
00225     int64_t max_packet_size;
00226     int compression_level;
00227     uint8_t *ptr, *top;
00228     uint8_t *crow_base = NULL, *crow_buf, *crow;
00229     uint8_t *progressive_buf = NULL;
00230     uint8_t *top_buf = NULL;
00231 
00232     *p = *pict;
00233     p->pict_type= AV_PICTURE_TYPE_I;
00234     p->key_frame= 1;
00235 
00236     is_progressive = !!(avctx->flags & CODEC_FLAG_INTERLACED_DCT);
00237     switch(avctx->pix_fmt) {
00238     case PIX_FMT_RGBA64BE:
00239         bit_depth = 16;
00240         color_type = PNG_COLOR_TYPE_RGB_ALPHA;
00241         break;
00242     case PIX_FMT_RGB48BE:
00243         bit_depth = 16;
00244         color_type = PNG_COLOR_TYPE_RGB;
00245         break;
00246     case PIX_FMT_RGBA:
00247         avctx->bits_per_coded_sample = 32;
00248         bit_depth = 8;
00249         color_type = PNG_COLOR_TYPE_RGB_ALPHA;
00250         break;
00251     case PIX_FMT_RGB24:
00252         avctx->bits_per_coded_sample = 24;
00253         bit_depth = 8;
00254         color_type = PNG_COLOR_TYPE_RGB;
00255         break;
00256     case PIX_FMT_GRAY16BE:
00257         bit_depth = 16;
00258         color_type = PNG_COLOR_TYPE_GRAY;
00259         break;
00260     case PIX_FMT_GRAY8:
00261         avctx->bits_per_coded_sample = 0x28;
00262         bit_depth = 8;
00263         color_type = PNG_COLOR_TYPE_GRAY;
00264         break;
00265     case PIX_FMT_GRAY8A:
00266         bit_depth = 8;
00267         color_type = PNG_COLOR_TYPE_GRAY_ALPHA;
00268         break;
00269     case PIX_FMT_MONOBLACK:
00270         avctx->bits_per_coded_sample =
00271         bit_depth = 1;
00272         color_type = PNG_COLOR_TYPE_GRAY;
00273         break;
00274     case PIX_FMT_PAL8:
00275         avctx->bits_per_coded_sample =
00276         bit_depth = 8;
00277         color_type = PNG_COLOR_TYPE_PALETTE;
00278         break;
00279     default:
00280         return -1;
00281     }
00282     bits_per_pixel = ff_png_get_nb_channels(color_type) * bit_depth;
00283     row_size = (avctx->width * bits_per_pixel + 7) >> 3;
00284 
00285     s->zstream.zalloc = ff_png_zalloc;
00286     s->zstream.zfree = ff_png_zfree;
00287     s->zstream.opaque = NULL;
00288     compression_level = avctx->compression_level == FF_COMPRESSION_DEFAULT ?
00289                             Z_DEFAULT_COMPRESSION :
00290                             av_clip(avctx->compression_level, 0, 9);
00291     ret = deflateInit2(&s->zstream, compression_level,
00292                        Z_DEFLATED, 15, 8, Z_DEFAULT_STRATEGY);
00293     if (ret != Z_OK)
00294         return -1;
00295 
00296     enc_row_size    = deflateBound(&s->zstream, row_size);
00297     max_packet_size = avctx->height * (int64_t)(enc_row_size +
00298                                        ((enc_row_size + IOBUF_SIZE - 1) / IOBUF_SIZE) * 12)
00299                       + FF_MIN_BUFFER_SIZE;
00300     if (max_packet_size > INT_MAX)
00301         return AVERROR(ENOMEM);
00302     if ((ret = ff_alloc_packet2(avctx, pkt, max_packet_size)) < 0)
00303         return ret;
00304 
00305     s->bytestream_start =
00306     s->bytestream       = pkt->data;
00307     s->bytestream_end   = pkt->data + pkt->size;
00308 
00309     crow_base = av_malloc((row_size + 32) << (s->filter_type == PNG_FILTER_VALUE_MIXED));
00310     if (!crow_base)
00311         goto fail;
00312     crow_buf = crow_base + 15; // pixel data should be aligned, but there's a control byte before it
00313     if (is_progressive) {
00314         progressive_buf = av_malloc(row_size + 1);
00315         if (!progressive_buf)
00316             goto fail;
00317     }
00318     if (is_progressive) {
00319         top_buf = av_malloc(row_size + 1);
00320         if (!top_buf)
00321             goto fail;
00322     }
00323 
00324     /* write png header */
00325     AV_WB64(s->bytestream, PNGSIG);
00326     s->bytestream += 8;
00327 
00328     AV_WB32(s->buf, avctx->width);
00329     AV_WB32(s->buf + 4, avctx->height);
00330     s->buf[8] = bit_depth;
00331     s->buf[9] = color_type;
00332     s->buf[10] = 0; /* compression type */
00333     s->buf[11] = 0; /* filter type */
00334     s->buf[12] = is_progressive; /* interlace type */
00335 
00336     png_write_chunk(&s->bytestream, MKTAG('I', 'H', 'D', 'R'), s->buf, 13);
00337 
00338     /* put the palette if needed */
00339     if (color_type == PNG_COLOR_TYPE_PALETTE) {
00340         int has_alpha, alpha, i;
00341         unsigned int v;
00342         uint32_t *palette;
00343         uint8_t *alpha_ptr;
00344 
00345         palette = (uint32_t *)p->data[1];
00346         ptr = s->buf;
00347         alpha_ptr = s->buf + 256 * 3;
00348         has_alpha = 0;
00349         for(i = 0; i < 256; i++) {
00350             v = palette[i];
00351             alpha = v >> 24;
00352             if (alpha != 0xff)
00353                 has_alpha = 1;
00354             *alpha_ptr++ = alpha;
00355             bytestream_put_be24(&ptr, v);
00356         }
00357         png_write_chunk(&s->bytestream, MKTAG('P', 'L', 'T', 'E'), s->buf, 256 * 3);
00358         if (has_alpha) {
00359             png_write_chunk(&s->bytestream, MKTAG('t', 'R', 'N', 'S'), s->buf + 256 * 3, 256);
00360         }
00361     }
00362 
00363     /* now put each row */
00364     s->zstream.avail_out = IOBUF_SIZE;
00365     s->zstream.next_out = s->buf;
00366     if (is_progressive) {
00367         int pass;
00368 
00369         for(pass = 0; pass < NB_PASSES; pass++) {
00370             /* NOTE: a pass is completely omited if no pixels would be
00371                output */
00372             pass_row_size = ff_png_pass_row_size(pass, bits_per_pixel, avctx->width);
00373             if (pass_row_size > 0) {
00374                 top = NULL;
00375                 for(y = 0; y < avctx->height; y++) {
00376                     if ((ff_png_pass_ymask[pass] << (y & 7)) & 0x80) {
00377                         ptr = p->data[0] + y * p->linesize[0];
00378                         FFSWAP(uint8_t*, progressive_buf, top_buf);
00379                         png_get_interlaced_row(progressive_buf, pass_row_size,
00380                                                bits_per_pixel, pass,
00381                                                ptr, avctx->width);
00382                         crow = png_choose_filter(s, crow_buf, progressive_buf, top, pass_row_size, bits_per_pixel>>3);
00383                         png_write_row(s, crow, pass_row_size + 1);
00384                         top = progressive_buf;
00385                     }
00386                 }
00387             }
00388         }
00389     } else {
00390         top = NULL;
00391         for(y = 0; y < avctx->height; y++) {
00392             ptr = p->data[0] + y * p->linesize[0];
00393             crow = png_choose_filter(s, crow_buf, ptr, top, row_size, bits_per_pixel>>3);
00394             png_write_row(s, crow, row_size + 1);
00395             top = ptr;
00396         }
00397     }
00398     /* compress last bytes */
00399     for(;;) {
00400         ret = deflate(&s->zstream, Z_FINISH);
00401         if (ret == Z_OK || ret == Z_STREAM_END) {
00402             len = IOBUF_SIZE - s->zstream.avail_out;
00403             if (len > 0 && s->bytestream_end - s->bytestream > len + 100) {
00404                 png_write_chunk(&s->bytestream, MKTAG('I', 'D', 'A', 'T'), s->buf, len);
00405             }
00406             s->zstream.avail_out = IOBUF_SIZE;
00407             s->zstream.next_out = s->buf;
00408             if (ret == Z_STREAM_END)
00409                 break;
00410         } else {
00411             goto fail;
00412         }
00413     }
00414     png_write_chunk(&s->bytestream, MKTAG('I', 'E', 'N', 'D'), NULL, 0);
00415 
00416     pkt->size   = s->bytestream - s->bytestream_start;
00417     pkt->flags |= AV_PKT_FLAG_KEY;
00418     *got_packet = 1;
00419     ret         = 0;
00420 
00421  the_end:
00422     av_free(crow_base);
00423     av_free(progressive_buf);
00424     av_free(top_buf);
00425     deflateEnd(&s->zstream);
00426     return ret;
00427  fail:
00428     ret = -1;
00429     goto the_end;
00430 }
00431 
00432 static av_cold int png_enc_init(AVCodecContext *avctx){
00433     PNGEncContext *s = avctx->priv_data;
00434 
00435     avcodec_get_frame_defaults(&s->picture);
00436     avctx->coded_frame= &s->picture;
00437     ff_dsputil_init(&s->dsp, avctx);
00438 
00439     s->filter_type = av_clip(avctx->prediction_method, PNG_FILTER_VALUE_NONE, PNG_FILTER_VALUE_MIXED);
00440     if(avctx->pix_fmt == PIX_FMT_MONOBLACK)
00441         s->filter_type = PNG_FILTER_VALUE_NONE;
00442 
00443     return 0;
00444 }
00445 
00446 AVCodec ff_png_encoder = {
00447     .name           = "png",
00448     .type           = AVMEDIA_TYPE_VIDEO,
00449     .id             = AV_CODEC_ID_PNG,
00450     .priv_data_size = sizeof(PNGEncContext),
00451     .init           = png_enc_init,
00452     .encode2        = encode_frame,
00453     .capabilities   = CODEC_CAP_FRAME_THREADS | CODEC_CAP_INTRA_ONLY,
00454     .pix_fmts       = (const enum PixelFormat[]){
00455         PIX_FMT_RGB24, PIX_FMT_RGBA,
00456         PIX_FMT_RGB48BE, PIX_FMT_RGBA64BE,
00457         PIX_FMT_PAL8,
00458         PIX_FMT_GRAY8, PIX_FMT_GRAY8A,
00459         PIX_FMT_GRAY16BE,
00460         PIX_FMT_MONOBLACK, PIX_FMT_NONE
00461     },
00462     .long_name      = NULL_IF_CONFIG_SMALL("PNG (Portable Network Graphics) image"),
00463 };