libavcodec : add decoder for Photoshop PSD image files
authorMartin Vignali <martin.vignali@gmail.com>
Thu, 24 Nov 2016 20:26:42 +0000 (21:26 +0100)
committerMichael Niedermayer <michael@niedermayer.cc>
Wed, 14 Dec 2016 02:21:27 +0000 (03:21 +0100)
Decode the Image Data Section (which contains merged pictures).
Support RGB/A and Grayscale/A in 8bits and 16 bits per channel.
Support uncompress and rle decompression in Image Data Section.

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Changelog
doc/general.texi
libavcodec/Makefile
libavcodec/allcodecs.c
libavcodec/avcodec.h
libavcodec/codec_desc.c
libavcodec/psd.c [new file with mode: 0644]

index 37fbb36..b36a631 100644 (file)
--- a/Changelog
+++ b/Changelog
@@ -8,6 +8,7 @@ version <next>:
 - premultiply video filter
 - Support for spherical videos
 - configure now fails if autodetect-libraries are requested but not found
+- PSD Decoder
 
 version 3.2:
 - libopenmpt demuxer
index 56b9e4d..9ea3ba3 100644 (file)
@@ -584,6 +584,8 @@ following image formats are supported:
 @item PNG          @tab X @tab X
 @item PPM          @tab X @tab X
     @tab Portable PixelMap image
+@item PSD          @tab   @tab X
+    @tab Photoshop
 @item PTX          @tab   @tab X
     @tab V.Flash PTX format
 @item SGI          @tab X @tab X
index 6dd294e..23e41dd 100644 (file)
@@ -462,6 +462,7 @@ OBJS-$(CONFIG_PRORES_LGPL_DECODER)     += proresdec_lgpl.o proresdsp.o proresdat
 OBJS-$(CONFIG_PRORES_ENCODER)          += proresenc_anatoliy.o
 OBJS-$(CONFIG_PRORES_AW_ENCODER)       += proresenc_anatoliy.o
 OBJS-$(CONFIG_PRORES_KS_ENCODER)       += proresenc_kostya.o proresdata.o
+OBJS-$(CONFIG_PSD_DECODER)             += psd.o
 OBJS-$(CONFIG_PTX_DECODER)             += ptx.o
 OBJS-$(CONFIG_QCELP_DECODER)           += qcelpdec.o                     \
                                           celp_filters.o acelp_vectors.o \
index ada9481..bbcecce 100644 (file)
@@ -287,6 +287,7 @@ void avcodec_register_all(void)
     REGISTER_ENCODER(PRORES_AW,         prores_aw);
     REGISTER_ENCODER(PRORES_KS,         prores_ks);
     REGISTER_DECODER(PRORES_LGPL,       prores_lgpl);
+    REGISTER_DECODER(PSD,               psd);
     REGISTER_DECODER(PTX,               ptx);
     REGISTER_DECODER(QDRAW,             qdraw);
     REGISTER_DECODER(QPEG,              qpeg);
index 02234ae..098debf 100644 (file)
@@ -411,6 +411,7 @@ enum AVCodecID {
     AV_CODEC_ID_MAGICYUV,
     AV_CODEC_ID_SHEERVIDEO,
     AV_CODEC_ID_YLC,
+    AV_CODEC_ID_PSD,
 
     /* various PCM "codecs" */
     AV_CODEC_ID_FIRST_AUDIO = 0x10000,     ///< A dummy id pointing at the start of audio codecs
index 9dbe2dc..29ffcb9 100644 (file)
@@ -1461,6 +1461,13 @@ static const AVCodecDescriptor codec_descriptors[] = {
         .props     = AV_CODEC_PROP_INTRA_ONLY | AV_CODEC_PROP_LOSSLESS,
     },
     {
+        .id        = AV_CODEC_ID_PSD,
+        .type      = AVMEDIA_TYPE_VIDEO,
+        .name      = "psd",
+        .long_name = NULL_IF_CONFIG_SMALL("Photoshop PSD file"),
+        .props     = AV_CODEC_PROP_INTRA_ONLY | AV_CODEC_PROP_LOSSLESS,
+    },
+    {
         .id        = AV_CODEC_ID_PTX,
         .type      = AVMEDIA_TYPE_VIDEO,
         .name      = "ptx",
diff --git a/libavcodec/psd.c b/libavcodec/psd.c
new file mode 100644 (file)
index 0000000..d5a4f52
--- /dev/null
@@ -0,0 +1,435 @@
+/*
+ * Photoshop (PSD) image decoder
+ * Copyright (c) 2016 Jokyo Images
+ *
+ * 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
+ */
+
+#include "bytestream.h"
+#include "internal.h"
+
+enum PsdCompr {
+    PSD_RAW,
+    PSD_RLE,
+    PSD_ZIP_WITHOUT_P,
+    PSD_ZIP_WITH_P,
+};
+
+enum PsdColorMode {
+    PSD_BITMAP,
+    PSD_GRAYSCALE,
+    PSD_INDEXED,
+    PSD_RGB,
+    PSD_CMYK,
+    PSD_MULTICHANNEL,
+    PSD_DUOTONE,
+    PSD_LAB,
+};
+
+typedef struct PSDContext {
+    AVClass *class;
+    AVFrame *picture;
+    AVCodecContext *avctx;
+    GetByteContext gb;
+
+    uint8_t * tmp;
+
+    uint16_t channel_count;
+    uint16_t channel_depth;
+
+    uint64_t uncompressed_size;
+    unsigned int pixel_size;/* 1 for 8 bits, 2 for 16 bits */
+    uint64_t line_size;/* length of src data (even width) */
+
+    int width;
+    int height;
+
+    enum PsdCompr compression;
+    enum PsdColorMode color_mode;
+} PSDContext;
+
+static int decode_header(PSDContext * s)
+{
+    int signature, version, color_mode, compression;
+    int64_t len_section;
+    int ret = 0;
+
+    if (bytestream2_get_bytes_left(&s->gb) < 30) {/* File header section + color map data section length */
+        av_log(s->avctx, AV_LOG_ERROR, "Header too short to parse.\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    signature = bytestream2_get_le32(&s->gb);
+    if (signature != MKTAG('8','B','P','S')) {
+        av_log(s->avctx, AV_LOG_ERROR, "Wrong signature %d.\n", signature);
+        return AVERROR_INVALIDDATA;
+    }
+
+    version = bytestream2_get_be16(&s->gb);
+    if (version != 1) {
+        av_log(s->avctx, AV_LOG_ERROR, "Wrong version %d.\n", version);
+        return AVERROR_INVALIDDATA;
+    }
+
+    bytestream2_skip(&s->gb, 6);/* reserved */
+
+    s->channel_count = bytestream2_get_be16(&s->gb);
+    if ((s->channel_count < 1) || (s->channel_count > 56)) {
+        av_log(s->avctx, AV_LOG_ERROR, "Invalid channel count %d.\n", s->channel_count);
+        return AVERROR_INVALIDDATA;
+    }
+
+    s->height = bytestream2_get_be32(&s->gb);
+
+    if ((s->height > 30000) && (s->avctx->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL)) {
+        av_log(s->avctx, AV_LOG_ERROR,
+               "Height > 30000 is experimental, add "
+               "'-strict %d' if you want to try to decode the picture.\n",
+               FF_COMPLIANCE_EXPERIMENTAL);
+        return AVERROR_EXPERIMENTAL;
+    }
+
+    s->width = bytestream2_get_be32(&s->gb);
+    if ((s->width > 30000) && (s->avctx->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL)) {
+        av_log(s->avctx, AV_LOG_ERROR,
+               "Width > 30000 is experimental, add "
+               "'-strict %d' if you want to try to decode the picture.\n",
+               FF_COMPLIANCE_EXPERIMENTAL);
+        return AVERROR_EXPERIMENTAL;
+    }
+
+    if ((ret = ff_set_dimensions(s->avctx, s->width, s->height)) < 0)
+        return ret;
+
+    s->channel_depth = bytestream2_get_be16(&s->gb);
+
+    color_mode = bytestream2_get_be16(&s->gb);
+    switch (color_mode) {
+    case 0:
+        s->color_mode = PSD_BITMAP;
+        break;
+    case 1:
+        s->color_mode = PSD_GRAYSCALE;
+        break;
+    case 2:
+        s->color_mode = PSD_INDEXED;
+        break;
+    case 3:
+        s->color_mode = PSD_RGB;
+        break;
+    case 4:
+        s->color_mode = PSD_CMYK;
+        break;
+    case 7:
+        s->color_mode = PSD_MULTICHANNEL;
+        break;
+    case 8:
+        s->color_mode = PSD_DUOTONE;
+        break;
+    case 9:
+        s->color_mode = PSD_LAB;
+        break;
+    default:
+        av_log(s->avctx, AV_LOG_ERROR, "Unknown color mode %d.\n", color_mode);
+        return AVERROR_INVALIDDATA;
+    }
+
+    /* color map data */
+    len_section = bytestream2_get_be32(&s->gb);
+    if (len_section < 0) {
+        av_log(s->avctx, AV_LOG_ERROR, "Negative size for color map data section.\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    if (bytestream2_get_bytes_left(&s->gb) < (len_section + 4)) { /* section and len next section */
+        av_log(s->avctx, AV_LOG_ERROR, "Incomplete file.\n");
+        return AVERROR_INVALIDDATA;
+    }
+    bytestream2_skip(&s->gb, len_section);
+
+    /* image ressources */
+    len_section = bytestream2_get_be32(&s->gb);
+    if (len_section < 0) {
+        av_log(s->avctx, AV_LOG_ERROR, "Negative size for image ressources section.\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    if (bytestream2_get_bytes_left(&s->gb) < (len_section + 4)) { /* section and len next section */
+        av_log(s->avctx, AV_LOG_ERROR, "Incomplete file.\n");
+        return AVERROR_INVALIDDATA;
+    }
+    bytestream2_skip(&s->gb, len_section);
+
+    /* layers and masks */
+    len_section = bytestream2_get_be32(&s->gb);
+    if (len_section < 0) {
+        av_log(s->avctx, AV_LOG_ERROR, "Negative size for layers and masks data section.\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    if (bytestream2_get_bytes_left(&s->gb) < len_section) {
+        av_log(s->avctx, AV_LOG_ERROR, "Incomplete file.\n");
+        return AVERROR_INVALIDDATA;
+    }
+    bytestream2_skip(&s->gb, len_section);
+
+    /* image section */
+    if (bytestream2_get_bytes_left(&s->gb) < 2) {
+        av_log(s->avctx, AV_LOG_ERROR, "File without image data section.\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    s->compression = bytestream2_get_be16(&s->gb);
+    switch (s->compression) {
+    case 0:
+    case 1:
+        break;
+    case 2:
+        avpriv_request_sample(s->avctx, "ZIP without predictor compression");
+        return AVERROR_PATCHWELCOME;
+        break;
+    case 3:
+        avpriv_request_sample(s->avctx, "ZIP with predictor compression");
+        return AVERROR_PATCHWELCOME;
+        break;
+    default:
+        av_log(s->avctx, AV_LOG_ERROR, "Unknown compression %d.\n", compression);
+        return AVERROR_INVALIDDATA;
+    }
+
+    return ret;
+}
+
+static int decode_rle(PSDContext * s){
+    unsigned int scanline_count;
+    unsigned int sl, count;
+    unsigned long target_index = 0;
+    unsigned int p;
+    int8_t rle_char;
+    unsigned int repeat_count;
+    uint8_t v;
+
+    scanline_count = s->height * s->channel_count;
+
+    /* scanline table */
+    if (bytestream2_get_bytes_left(&s->gb) < scanline_count * 2) {
+        av_log(s->avctx, AV_LOG_ERROR, "Not enough data for rle scanline table.\n");
+        return AVERROR_INVALIDDATA;
+    }
+    bytestream2_skip(&s->gb, scanline_count * 2);/* size of each scanline */
+
+    /* decode rle data scanline by scanline */
+    for (sl = 0; sl < scanline_count; sl++) {
+        count = 0;
+
+        while (count < s->line_size) {
+            rle_char = bytestream2_get_byte(&s->gb);
+
+            if (rle_char <= 0) {/* byte repeat */
+                repeat_count = rle_char * -1;
+
+                if (bytestream2_get_bytes_left(&s->gb) < 1) {
+                    av_log(s->avctx, AV_LOG_ERROR, "Not enough data for rle scanline.\n");
+                    return AVERROR_INVALIDDATA;
+                }
+
+                if (target_index + repeat_count >= s->uncompressed_size) {
+                    av_log(s->avctx, AV_LOG_ERROR, "Invalid rle char.\n");
+                    return AVERROR_INVALIDDATA;
+                }
+
+                v = bytestream2_get_byte(&s->gb);
+                for (p = 0; p <= repeat_count; p++) {
+                    s->tmp[target_index++] = v;
+                }
+                count += repeat_count + 1;
+            } else {
+                if (bytestream2_get_bytes_left(&s->gb) < rle_char) {
+                    av_log(s->avctx, AV_LOG_ERROR, "Not enough data for rle scanline.\n");
+                    return AVERROR_INVALIDDATA;
+                }
+
+                if (target_index + rle_char >= s->uncompressed_size) {
+                    av_log(s->avctx, AV_LOG_ERROR, "Invalid rle char.\n");
+                    return AVERROR_INVALIDDATA;
+                }
+
+                for (p = 0; p <= rle_char; p++) {
+                    v = bytestream2_get_byte(&s->gb);
+                    s->tmp[target_index++] = v;
+                }
+                count += rle_char + 1;
+            }
+        }
+    }
+
+    return 0;
+}
+
+static int decode_frame(AVCodecContext *avctx, void *data,
+                        int *got_frame, AVPacket *avpkt)
+{
+    int ret;
+    uint8_t *ptr;
+    const uint8_t *ptr_data;
+    int index_out, c, y, x, p;
+    uint8_t eq_channel[4] = {2,0,1,3};/* RGBA -> GBRA channel order */
+    uint8_t plane_number;
+
+    AVFrame *picture = data;
+
+    PSDContext *s = avctx->priv_data;
+    s->avctx     = avctx;
+    s->channel_count = 0;
+    s->channel_depth = 0;
+    s->tmp           = NULL;
+    s->line_size     = 0;
+
+    bytestream2_init(&s->gb, avpkt->data, avpkt->size);
+
+    if ((ret = decode_header(s)) < 0)
+        return ret;
+
+    s->pixel_size = s->channel_depth >> 3;/* in byte */
+    s->line_size = s->width * s->pixel_size;
+    s->uncompressed_size = s->line_size * s->height * s->channel_count;
+
+    switch (s->color_mode) {
+    case PSD_RGB:
+        if (s->channel_count == 3) {
+            if (s->channel_depth == 8) {
+                avctx->pix_fmt = AV_PIX_FMT_GBRP;
+            } else if (s->channel_depth == 16) {
+                avctx->pix_fmt = AV_PIX_FMT_GBRP16BE;
+            } else {
+                avpriv_report_missing_feature(avctx, "channel depth %d for rgb", s->channel_depth);
+                return AVERROR_PATCHWELCOME;
+            }
+        } else if (s->channel_count == 4) {
+            if (s->channel_depth == 8) {
+                avctx->pix_fmt = AV_PIX_FMT_GBRAP;
+            } else if (s->channel_depth == 16) {
+                avctx->pix_fmt = AV_PIX_FMT_GBRAP16BE;
+            } else {
+                avpriv_report_missing_feature(avctx, "channel depth %d for rgb", s->channel_depth);
+                return AVERROR_PATCHWELCOME;
+            }
+        } else {
+            avpriv_report_missing_feature(avctx, "channel count %d for rgb", s->channel_count);
+            return AVERROR_PATCHWELCOME;
+        }
+        break;
+    case PSD_GRAYSCALE:
+        if (s->channel_count == 1) {
+            if (s->channel_depth == 8) {
+                avctx->pix_fmt = AV_PIX_FMT_GRAY8;
+            } else if (s->channel_depth == 16) {
+                avctx->pix_fmt = AV_PIX_FMT_GRAY16BE;
+            } else {
+                avpriv_report_missing_feature(avctx, "channel depth %d for grayscale", s->channel_depth);
+                return AVERROR_PATCHWELCOME;
+            }
+        } else if (s->channel_count == 2) {
+            if (s->channel_depth == 8) {
+                avctx->pix_fmt = AV_PIX_FMT_YA8;
+            } else if (s->channel_depth == 16) {
+                avctx->pix_fmt = AV_PIX_FMT_YA16BE;
+            } else {
+                avpriv_report_missing_feature(avctx, "channel depth %d for grayscale", s->channel_depth);
+                return AVERROR_PATCHWELCOME;
+            }
+        } else {
+            avpriv_report_missing_feature(avctx, "channel count %d for grayscale", s->channel_count);
+            return AVERROR_PATCHWELCOME;
+        }
+        break;
+    default:
+        avpriv_report_missing_feature(avctx, "color mode %d", s->color_mode);
+        return AVERROR_PATCHWELCOME;
+    }
+
+    if ((ret = ff_get_buffer(avctx, picture, 0)) < 0)
+        return ret;
+
+    /* decode picture if need */
+    if (s->compression == PSD_RLE) {
+        s->tmp = av_malloc(s->uncompressed_size);
+        if (!s->tmp)
+            return AVERROR(ENOMEM);
+
+        ret = decode_rle(s);
+
+        if (ret < 0) {
+            av_freep(&s->tmp);
+            return ret;
+        }
+
+        ptr_data = s->tmp;
+    } else {
+        if (bytestream2_get_bytes_left(&s->gb) < s->uncompressed_size) {
+            av_log(s->avctx, AV_LOG_ERROR, "Not enough data for raw image data section.\n");
+            return AVERROR_INVALIDDATA;
+        }
+        ptr_data = s->gb.buffer;
+    }
+
+    /* Store data */
+    if ((avctx->pix_fmt == AV_PIX_FMT_YA8)||(avctx->pix_fmt == AV_PIX_FMT_YA16BE)){/* Interleaved */
+        ptr = picture->data[0];
+        for (c = 0; c < s->channel_count; c++) {
+            for (y = 0; y < s->height; y++) {
+                for (x = 0; x < s->width; x++) {
+                    index_out = y * picture->linesize[0] + x * s->channel_count * s->pixel_size + c * s->pixel_size;
+                    for (p = 0; p < s->pixel_size; p++) {
+                        ptr[index_out + p] = *ptr_data;
+                        ptr_data ++;
+                    }
+                }
+            }
+        }
+    } else {/* Planar */
+        if (s->channel_count == 1)/* gray 8 or gray 16be */
+            eq_channel[0] = 0;/* assign first channel, to first plane */
+
+        for (c = 0; c < s->channel_count; c++) {
+            plane_number = eq_channel[c];
+            ptr = picture->data[plane_number];/* get the right plane */
+            for (y = 0; y < s->height; y++) {
+                memcpy(ptr, ptr_data, s->width * s->pixel_size);
+                ptr += picture->linesize[plane_number];
+                ptr_data += s->width * s->pixel_size;
+            }
+        }
+    }
+
+    av_freep(&s->tmp);
+
+    picture->pict_type = AV_PICTURE_TYPE_I;
+    *got_frame = 1;
+
+    return avpkt->size;
+}
+
+AVCodec ff_psd_decoder = {
+    .name             = "psd",
+    .long_name        = NULL_IF_CONFIG_SMALL("Photoshop PSD file"),
+    .type             = AVMEDIA_TYPE_VIDEO,
+    .id               = AV_CODEC_ID_PSD,
+    .priv_data_size   = sizeof(PSDContext),
+    .decode           = decode_frame,
+    .capabilities     = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS,
+};