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