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