mirror of https://github.com/FFmpeg/FFmpeg.git
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
752 lines
28 KiB
752 lines
28 KiB
/* |
|
* FLI/FLC Animation Video Decoder |
|
* Copyright (C) 2003, 2004 the ffmpeg project |
|
* |
|
* 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 |
|
* |
|
*/ |
|
|
|
/** |
|
* @file flic.c |
|
* Autodesk Animator FLI/FLC Video Decoder |
|
* by Mike Melanson (melanson@pcisys.net) |
|
* for more information on the .fli/.flc file format and all of its many |
|
* variations, visit: |
|
* http://www.compuphase.com/flic.htm |
|
* |
|
* This decoder outputs PAL8/RGB555/RGB565 and maybe one day RGB24 |
|
* colorspace data, depending on the FLC. To use this decoder, be |
|
* sure that your demuxer sends the FLI file header to the decoder via |
|
* the extradata chunk in AVCodecContext. The chunk should be 128 bytes |
|
* large. The only exception is for FLI files from the game "Magic Carpet", |
|
* in which the header is only 12 bytes. |
|
*/ |
|
|
|
#include <stdio.h> |
|
#include <stdlib.h> |
|
#include <string.h> |
|
#include <unistd.h> |
|
|
|
#include "avcodec.h" |
|
#include "bswap.h" |
|
|
|
#define FLI_256_COLOR 4 |
|
#define FLI_DELTA 7 |
|
#define FLI_COLOR 11 |
|
#define FLI_LC 12 |
|
#define FLI_BLACK 13 |
|
#define FLI_BRUN 15 |
|
#define FLI_COPY 16 |
|
#define FLI_MINI 18 |
|
#define FLI_DTA_BRUN 25 |
|
#define FLI_DTA_COPY 26 |
|
#define FLI_DTA_LC 27 |
|
|
|
#define FLI_TYPE_CODE (0xAF11) |
|
#define FLC_FLX_TYPE_CODE (0xAF12) |
|
#define FLC_DTA_TYPE_CODE (0xAF44) /* Marks an "Extended FLC" comes from Dave's Targa Animator (DTA) */ |
|
#define FLC_MAGIC_CARPET_SYNTHETIC_TYPE_CODE (0xAF13) |
|
|
|
#define CHECK_PIXEL_PTR(n) \ |
|
if (pixel_ptr + n > pixel_limit) { \ |
|
av_log (s->avctx, AV_LOG_INFO, "Problem: pixel_ptr >= pixel_limit (%d >= %d)\n", \ |
|
pixel_ptr + n, pixel_limit); \ |
|
return -1; \ |
|
} \ |
|
|
|
typedef struct FlicDecodeContext { |
|
AVCodecContext *avctx; |
|
AVFrame frame; |
|
|
|
unsigned int palette[256]; |
|
int new_palette; |
|
int fli_type; /* either 0xAF11 or 0xAF12, affects palette resolution */ |
|
} FlicDecodeContext; |
|
|
|
static int flic_decode_init(AVCodecContext *avctx) |
|
{ |
|
FlicDecodeContext *s = avctx->priv_data; |
|
unsigned char *fli_header = (unsigned char *)avctx->extradata; |
|
int depth; |
|
|
|
s->avctx = avctx; |
|
|
|
s->fli_type = AV_RL16(&fli_header[4]); /* Might be overridden if a Magic Carpet FLC */ |
|
depth = AV_RL16(&fli_header[12]); |
|
|
|
if (depth == 0) { |
|
depth = 8; /* Some FLC generators set depth to zero, when they mean 8Bpp. Fix up here */ |
|
} |
|
|
|
if (s->avctx->extradata_size == 12) { |
|
/* special case for magic carpet FLIs */ |
|
s->fli_type = FLC_MAGIC_CARPET_SYNTHETIC_TYPE_CODE; |
|
} else if (s->avctx->extradata_size != 128) { |
|
av_log(avctx, AV_LOG_ERROR, "Expected extradata of 12 or 128 bytes\n"); |
|
return -1; |
|
} |
|
|
|
if ((s->fli_type == FLC_FLX_TYPE_CODE) && (depth == 16)) { |
|
depth = 15; /* Original Autodesk FLX's say the depth is 16Bpp when it is really 15Bpp */ |
|
} |
|
|
|
switch (depth) { |
|
case 8 : avctx->pix_fmt = PIX_FMT_PAL8; break; |
|
case 15 : avctx->pix_fmt = PIX_FMT_RGB555; break; |
|
case 16 : avctx->pix_fmt = PIX_FMT_RGB565; break; |
|
case 24 : avctx->pix_fmt = PIX_FMT_BGR24; /* Supposedly BGR, but havent any files to test with */ |
|
av_log(avctx, AV_LOG_ERROR, "24Bpp FLC/FLX is unsupported due to no test files.\n"); |
|
return -1; |
|
break; |
|
default : |
|
av_log(avctx, AV_LOG_ERROR, "Unknown FLC/FLX depth of %d Bpp is unsupported.\n",depth); |
|
return -1; |
|
} |
|
|
|
s->frame.data[0] = NULL; |
|
s->new_palette = 0; |
|
|
|
return 0; |
|
} |
|
|
|
static int flic_decode_frame_8BPP(AVCodecContext *avctx, |
|
void *data, int *data_size, |
|
uint8_t *buf, int buf_size) |
|
{ |
|
FlicDecodeContext *s = avctx->priv_data; |
|
|
|
int stream_ptr = 0; |
|
int stream_ptr_after_color_chunk; |
|
int pixel_ptr; |
|
int palette_ptr; |
|
unsigned char palette_idx1; |
|
unsigned char palette_idx2; |
|
|
|
unsigned int frame_size; |
|
int num_chunks; |
|
|
|
unsigned int chunk_size; |
|
int chunk_type; |
|
|
|
int i, j; |
|
|
|
int color_packets; |
|
int color_changes; |
|
int color_shift; |
|
unsigned char r, g, b; |
|
|
|
int lines; |
|
int compressed_lines; |
|
int starting_line; |
|
signed short line_packets; |
|
int y_ptr; |
|
int byte_run; |
|
int pixel_skip; |
|
int pixel_countdown; |
|
unsigned char *pixels; |
|
int pixel_limit; |
|
|
|
s->frame.reference = 1; |
|
s->frame.buffer_hints = FF_BUFFER_HINTS_VALID | FF_BUFFER_HINTS_PRESERVE | FF_BUFFER_HINTS_REUSABLE; |
|
if (avctx->reget_buffer(avctx, &s->frame) < 0) { |
|
av_log(avctx, AV_LOG_ERROR, "reget_buffer() failed\n"); |
|
return -1; |
|
} |
|
|
|
pixels = s->frame.data[0]; |
|
pixel_limit = s->avctx->height * s->frame.linesize[0]; |
|
|
|
frame_size = AV_RL32(&buf[stream_ptr]); |
|
stream_ptr += 6; /* skip the magic number */ |
|
num_chunks = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 10; /* skip padding */ |
|
|
|
frame_size -= 16; |
|
|
|
/* iterate through the chunks */ |
|
while ((frame_size > 0) && (num_chunks > 0)) { |
|
chunk_size = AV_RL32(&buf[stream_ptr]); |
|
stream_ptr += 4; |
|
chunk_type = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 2; |
|
|
|
switch (chunk_type) { |
|
case FLI_256_COLOR: |
|
case FLI_COLOR: |
|
stream_ptr_after_color_chunk = stream_ptr + chunk_size - 6; |
|
|
|
/* check special case: If this file is from the Magic Carpet |
|
* game and uses 6-bit colors even though it reports 256-color |
|
* chunks in a 0xAF12-type file (fli_type is set to 0xAF13 during |
|
* initialization) */ |
|
if ((chunk_type == FLI_256_COLOR) && (s->fli_type != FLC_MAGIC_CARPET_SYNTHETIC_TYPE_CODE)) |
|
color_shift = 0; |
|
else |
|
color_shift = 2; |
|
/* set up the palette */ |
|
color_packets = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 2; |
|
palette_ptr = 0; |
|
for (i = 0; i < color_packets; i++) { |
|
/* first byte is how many colors to skip */ |
|
palette_ptr += buf[stream_ptr++]; |
|
|
|
/* next byte indicates how many entries to change */ |
|
color_changes = buf[stream_ptr++]; |
|
|
|
/* if there are 0 color changes, there are actually 256 */ |
|
if (color_changes == 0) |
|
color_changes = 256; |
|
|
|
for (j = 0; j < color_changes; j++) { |
|
unsigned int entry; |
|
|
|
/* wrap around, for good measure */ |
|
if ((unsigned)palette_ptr >= 256) |
|
palette_ptr = 0; |
|
|
|
r = buf[stream_ptr++] << color_shift; |
|
g = buf[stream_ptr++] << color_shift; |
|
b = buf[stream_ptr++] << color_shift; |
|
entry = (r << 16) | (g << 8) | b; |
|
if (s->palette[palette_ptr] != entry) |
|
s->new_palette = 1; |
|
s->palette[palette_ptr++] = entry; |
|
} |
|
} |
|
|
|
/* color chunks sometimes have weird 16-bit alignment issues; |
|
* therefore, take the hardline approach and set the stream_ptr |
|
* to the value calculated w.r.t. the size specified by the color |
|
* chunk header */ |
|
stream_ptr = stream_ptr_after_color_chunk; |
|
|
|
break; |
|
|
|
case FLI_DELTA: |
|
y_ptr = 0; |
|
compressed_lines = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 2; |
|
while (compressed_lines > 0) { |
|
line_packets = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 2; |
|
if ((line_packets & 0xC000) == 0xC000) { |
|
// line skip opcode |
|
line_packets = -line_packets; |
|
y_ptr += line_packets * s->frame.linesize[0]; |
|
} else if ((line_packets & 0xC000) == 0x4000) { |
|
av_log(avctx, AV_LOG_ERROR, "Undefined opcode (%x) in DELTA_FLI\n", line_packets); |
|
} else if ((line_packets & 0xC000) == 0x8000) { |
|
// "last byte" opcode |
|
pixels[y_ptr + s->frame.linesize[0] - 1] = line_packets & 0xff; |
|
} else { |
|
compressed_lines--; |
|
pixel_ptr = y_ptr; |
|
pixel_countdown = s->avctx->width; |
|
for (i = 0; i < line_packets; i++) { |
|
/* account for the skip bytes */ |
|
pixel_skip = buf[stream_ptr++]; |
|
pixel_ptr += pixel_skip; |
|
pixel_countdown -= pixel_skip; |
|
byte_run = (signed char)(buf[stream_ptr++]); |
|
if (byte_run < 0) { |
|
byte_run = -byte_run; |
|
palette_idx1 = buf[stream_ptr++]; |
|
palette_idx2 = buf[stream_ptr++]; |
|
CHECK_PIXEL_PTR(byte_run); |
|
for (j = 0; j < byte_run; j++, pixel_countdown -= 2) { |
|
pixels[pixel_ptr++] = palette_idx1; |
|
pixels[pixel_ptr++] = palette_idx2; |
|
} |
|
} else { |
|
CHECK_PIXEL_PTR(byte_run * 2); |
|
for (j = 0; j < byte_run * 2; j++, pixel_countdown--) { |
|
palette_idx1 = buf[stream_ptr++]; |
|
pixels[pixel_ptr++] = palette_idx1; |
|
} |
|
} |
|
} |
|
|
|
y_ptr += s->frame.linesize[0]; |
|
} |
|
} |
|
break; |
|
|
|
case FLI_LC: |
|
/* line compressed */ |
|
starting_line = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 2; |
|
y_ptr = 0; |
|
y_ptr += starting_line * s->frame.linesize[0]; |
|
|
|
compressed_lines = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 2; |
|
while (compressed_lines > 0) { |
|
pixel_ptr = y_ptr; |
|
pixel_countdown = s->avctx->width; |
|
line_packets = buf[stream_ptr++]; |
|
if (line_packets > 0) { |
|
for (i = 0; i < line_packets; i++) { |
|
/* account for the skip bytes */ |
|
pixel_skip = buf[stream_ptr++]; |
|
pixel_ptr += pixel_skip; |
|
pixel_countdown -= pixel_skip; |
|
byte_run = (signed char)(buf[stream_ptr++]); |
|
if (byte_run > 0) { |
|
CHECK_PIXEL_PTR(byte_run); |
|
for (j = 0; j < byte_run; j++, pixel_countdown--) { |
|
palette_idx1 = buf[stream_ptr++]; |
|
pixels[pixel_ptr++] = palette_idx1; |
|
} |
|
} else if (byte_run < 0) { |
|
byte_run = -byte_run; |
|
palette_idx1 = buf[stream_ptr++]; |
|
CHECK_PIXEL_PTR(byte_run); |
|
for (j = 0; j < byte_run; j++, pixel_countdown--) { |
|
pixels[pixel_ptr++] = palette_idx1; |
|
} |
|
} |
|
} |
|
} |
|
|
|
y_ptr += s->frame.linesize[0]; |
|
compressed_lines--; |
|
} |
|
break; |
|
|
|
case FLI_BLACK: |
|
/* set the whole frame to color 0 (which is usually black) */ |
|
memset(pixels, 0, |
|
s->frame.linesize[0] * s->avctx->height); |
|
break; |
|
|
|
case FLI_BRUN: |
|
/* Byte run compression: This chunk type only occurs in the first |
|
* FLI frame and it will update the entire frame. */ |
|
y_ptr = 0; |
|
for (lines = 0; lines < s->avctx->height; lines++) { |
|
pixel_ptr = y_ptr; |
|
/* disregard the line packets; instead, iterate through all |
|
* pixels on a row */ |
|
stream_ptr++; |
|
pixel_countdown = s->avctx->width; |
|
while (pixel_countdown > 0) { |
|
byte_run = (signed char)(buf[stream_ptr++]); |
|
if (byte_run > 0) { |
|
palette_idx1 = buf[stream_ptr++]; |
|
CHECK_PIXEL_PTR(byte_run); |
|
for (j = 0; j < byte_run; j++) { |
|
pixels[pixel_ptr++] = palette_idx1; |
|
pixel_countdown--; |
|
if (pixel_countdown < 0) |
|
av_log(avctx, AV_LOG_ERROR, "pixel_countdown < 0 (%d) at line %d\n", |
|
pixel_countdown, lines); |
|
} |
|
} else { /* copy bytes if byte_run < 0 */ |
|
byte_run = -byte_run; |
|
CHECK_PIXEL_PTR(byte_run); |
|
for (j = 0; j < byte_run; j++) { |
|
palette_idx1 = buf[stream_ptr++]; |
|
pixels[pixel_ptr++] = palette_idx1; |
|
pixel_countdown--; |
|
if (pixel_countdown < 0) |
|
av_log(avctx, AV_LOG_ERROR, "pixel_countdown < 0 (%d) at line %d\n", |
|
pixel_countdown, lines); |
|
} |
|
} |
|
} |
|
|
|
y_ptr += s->frame.linesize[0]; |
|
} |
|
break; |
|
|
|
case FLI_COPY: |
|
/* copy the chunk (uncompressed frame) */ |
|
if (chunk_size - 6 > s->avctx->width * s->avctx->height) { |
|
av_log(avctx, AV_LOG_ERROR, "In chunk FLI_COPY : source data (%d bytes) " \ |
|
"bigger than image, skipping chunk\n", chunk_size - 6); |
|
stream_ptr += chunk_size - 6; |
|
} else { |
|
for (y_ptr = 0; y_ptr < s->frame.linesize[0] * s->avctx->height; |
|
y_ptr += s->frame.linesize[0]) { |
|
memcpy(&pixels[y_ptr], &buf[stream_ptr], |
|
s->avctx->width); |
|
stream_ptr += s->avctx->width; |
|
} |
|
} |
|
break; |
|
|
|
case FLI_MINI: |
|
/* some sort of a thumbnail? disregard this chunk... */ |
|
stream_ptr += chunk_size - 6; |
|
break; |
|
|
|
default: |
|
av_log(avctx, AV_LOG_ERROR, "Unrecognized chunk type: %d\n", chunk_type); |
|
break; |
|
} |
|
|
|
frame_size -= chunk_size; |
|
num_chunks--; |
|
} |
|
|
|
/* by the end of the chunk, the stream ptr should equal the frame |
|
* size (minus 1, possibly); if it doesn't, issue a warning */ |
|
if ((stream_ptr != buf_size) && (stream_ptr != buf_size - 1)) |
|
av_log(avctx, AV_LOG_ERROR, "Processed FLI chunk where chunk size = %d " \ |
|
"and final chunk ptr = %d\n", buf_size, stream_ptr); |
|
|
|
/* make the palette available on the way out */ |
|
memcpy(s->frame.data[1], s->palette, AVPALETTE_SIZE); |
|
if (s->new_palette) { |
|
s->frame.palette_has_changed = 1; |
|
s->new_palette = 0; |
|
} |
|
|
|
*data_size=sizeof(AVFrame); |
|
*(AVFrame*)data = s->frame; |
|
|
|
return buf_size; |
|
} |
|
|
|
static int flic_decode_frame_15_16BPP(AVCodecContext *avctx, |
|
void *data, int *data_size, |
|
uint8_t *buf, int buf_size) |
|
{ |
|
/* Note, the only difference between the 15Bpp and 16Bpp */ |
|
/* Format is the pixel format, the packets are processed the same. */ |
|
FlicDecodeContext *s = avctx->priv_data; |
|
|
|
int stream_ptr = 0; |
|
int pixel_ptr; |
|
unsigned char palette_idx1; |
|
|
|
unsigned int frame_size; |
|
int num_chunks; |
|
|
|
unsigned int chunk_size; |
|
int chunk_type; |
|
|
|
int i, j; |
|
|
|
int lines; |
|
int compressed_lines; |
|
signed short line_packets; |
|
int y_ptr; |
|
int byte_run; |
|
int pixel_skip; |
|
int pixel_countdown; |
|
unsigned char *pixels; |
|
int pixel; |
|
int pixel_limit; |
|
|
|
s->frame.reference = 1; |
|
s->frame.buffer_hints = FF_BUFFER_HINTS_VALID | FF_BUFFER_HINTS_PRESERVE | FF_BUFFER_HINTS_REUSABLE; |
|
if (avctx->reget_buffer(avctx, &s->frame) < 0) { |
|
av_log(avctx, AV_LOG_ERROR, "reget_buffer() failed\n"); |
|
return -1; |
|
} |
|
|
|
pixels = s->frame.data[0]; |
|
pixel_limit = s->avctx->height * s->frame.linesize[0]; |
|
|
|
frame_size = AV_RL32(&buf[stream_ptr]); |
|
stream_ptr += 6; /* skip the magic number */ |
|
num_chunks = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 10; /* skip padding */ |
|
|
|
frame_size -= 16; |
|
|
|
/* iterate through the chunks */ |
|
while ((frame_size > 0) && (num_chunks > 0)) { |
|
chunk_size = AV_RL32(&buf[stream_ptr]); |
|
stream_ptr += 4; |
|
chunk_type = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 2; |
|
|
|
switch (chunk_type) { |
|
case FLI_256_COLOR: |
|
case FLI_COLOR: |
|
/* For some reason, it seems that non-paletised flics do include one of these */ |
|
/* chunks in their first frame. Why i do not know, it seems rather extraneous */ |
|
/* av_log(avctx, AV_LOG_ERROR, "Unexpected Palette chunk %d in non-paletised FLC\n",chunk_type);*/ |
|
stream_ptr = stream_ptr + chunk_size - 6; |
|
break; |
|
|
|
case FLI_DELTA: |
|
case FLI_DTA_LC: |
|
y_ptr = 0; |
|
compressed_lines = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 2; |
|
while (compressed_lines > 0) { |
|
line_packets = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 2; |
|
if (line_packets < 0) { |
|
line_packets = -line_packets; |
|
y_ptr += line_packets * s->frame.linesize[0]; |
|
} else { |
|
compressed_lines--; |
|
pixel_ptr = y_ptr; |
|
pixel_countdown = s->avctx->width; |
|
for (i = 0; i < line_packets; i++) { |
|
/* account for the skip bytes */ |
|
pixel_skip = buf[stream_ptr++]; |
|
pixel_ptr += (pixel_skip*2); /* Pixel is 2 bytes wide */ |
|
pixel_countdown -= pixel_skip; |
|
byte_run = (signed char)(buf[stream_ptr++]); |
|
if (byte_run < 0) { |
|
byte_run = -byte_run; |
|
pixel = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 2; |
|
CHECK_PIXEL_PTR(byte_run); |
|
for (j = 0; j < byte_run; j++, pixel_countdown -= 2) { |
|
*((signed short*)(&pixels[pixel_ptr])) = pixel; |
|
pixel_ptr += 2; |
|
} |
|
} else { |
|
CHECK_PIXEL_PTR(byte_run); |
|
for (j = 0; j < byte_run; j++, pixel_countdown--) { |
|
*((signed short*)(&pixels[pixel_ptr])) = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 2; |
|
pixel_ptr += 2; |
|
} |
|
} |
|
} |
|
|
|
y_ptr += s->frame.linesize[0]; |
|
} |
|
} |
|
break; |
|
|
|
case FLI_LC: |
|
av_log(avctx, AV_LOG_ERROR, "Unexpected FLI_LC chunk in non-paletised FLC\n"); |
|
stream_ptr = stream_ptr + chunk_size - 6; |
|
break; |
|
|
|
case FLI_BLACK: |
|
/* set the whole frame to 0x0000 which is black in both 15Bpp and 16Bpp modes. */ |
|
memset(pixels, 0x0000, |
|
s->frame.linesize[0] * s->avctx->height); |
|
break; |
|
|
|
case FLI_BRUN: |
|
y_ptr = 0; |
|
for (lines = 0; lines < s->avctx->height; lines++) { |
|
pixel_ptr = y_ptr; |
|
/* disregard the line packets; instead, iterate through all |
|
* pixels on a row */ |
|
stream_ptr++; |
|
pixel_countdown = (s->avctx->width * 2); |
|
|
|
while (pixel_countdown > 0) { |
|
byte_run = (signed char)(buf[stream_ptr++]); |
|
if (byte_run > 0) { |
|
palette_idx1 = buf[stream_ptr++]; |
|
CHECK_PIXEL_PTR(byte_run); |
|
for (j = 0; j < byte_run; j++) { |
|
pixels[pixel_ptr++] = palette_idx1; |
|
pixel_countdown--; |
|
if (pixel_countdown < 0) |
|
av_log(avctx, AV_LOG_ERROR, "pixel_countdown < 0 (%d) (linea%d)\n", |
|
pixel_countdown, lines); |
|
} |
|
} else { /* copy bytes if byte_run < 0 */ |
|
byte_run = -byte_run; |
|
CHECK_PIXEL_PTR(byte_run); |
|
for (j = 0; j < byte_run; j++) { |
|
palette_idx1 = buf[stream_ptr++]; |
|
pixels[pixel_ptr++] = palette_idx1; |
|
pixel_countdown--; |
|
if (pixel_countdown < 0) |
|
av_log(avctx, AV_LOG_ERROR, "pixel_countdown < 0 (%d) at line %d\n", |
|
pixel_countdown, lines); |
|
} |
|
} |
|
} |
|
|
|
/* Now FLX is strange, in that it is "byte" as opposed to "pixel" run length compressed. |
|
* This does not give us any good oportunity to perform word endian conversion |
|
* during decompression. So if it is required (i.e., this is not a LE target, we do |
|
* a second pass over the line here, swapping the bytes. |
|
*/ |
|
pixel = 0xFF00; |
|
if (0xFF00 != AV_RL16(&pixel)) /* Check if it is not an LE target */ |
|
{ |
|
pixel_ptr = y_ptr; |
|
pixel_countdown = s->avctx->width; |
|
while (pixel_countdown > 0) { |
|
*((signed short*)(&pixels[pixel_ptr])) = AV_RL16(&buf[pixel_ptr]); |
|
pixel_ptr += 2; |
|
} |
|
} |
|
y_ptr += s->frame.linesize[0]; |
|
} |
|
break; |
|
|
|
case FLI_DTA_BRUN: |
|
y_ptr = 0; |
|
for (lines = 0; lines < s->avctx->height; lines++) { |
|
pixel_ptr = y_ptr; |
|
/* disregard the line packets; instead, iterate through all |
|
* pixels on a row */ |
|
stream_ptr++; |
|
pixel_countdown = s->avctx->width; /* Width is in pixels, not bytes */ |
|
|
|
while (pixel_countdown > 0) { |
|
byte_run = (signed char)(buf[stream_ptr++]); |
|
if (byte_run > 0) { |
|
pixel = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 2; |
|
CHECK_PIXEL_PTR(byte_run); |
|
for (j = 0; j < byte_run; j++) { |
|
*((signed short*)(&pixels[pixel_ptr])) = pixel; |
|
pixel_ptr += 2; |
|
pixel_countdown--; |
|
if (pixel_countdown < 0) |
|
av_log(avctx, AV_LOG_ERROR, "pixel_countdown < 0 (%d)\n", |
|
pixel_countdown); |
|
} |
|
} else { /* copy pixels if byte_run < 0 */ |
|
byte_run = -byte_run; |
|
CHECK_PIXEL_PTR(byte_run); |
|
for (j = 0; j < byte_run; j++) { |
|
*((signed short*)(&pixels[pixel_ptr])) = AV_RL16(&buf[stream_ptr]); |
|
stream_ptr += 2; |
|
pixel_ptr += 2; |
|
pixel_countdown--; |
|
if (pixel_countdown < 0) |
|
av_log(avctx, AV_LOG_ERROR, "pixel_countdown < 0 (%d)\n", |
|
pixel_countdown); |
|
} |
|
} |
|
} |
|
|
|
y_ptr += s->frame.linesize[0]; |
|
} |
|
break; |
|
|
|
case FLI_COPY: |
|
case FLI_DTA_COPY: |
|
/* copy the chunk (uncompressed frame) */ |
|
if (chunk_size - 6 > (unsigned int)(s->avctx->width * s->avctx->height)*2) { |
|
av_log(avctx, AV_LOG_ERROR, "In chunk FLI_COPY : source data (%d bytes) " \ |
|
"bigger than image, skipping chunk\n", chunk_size - 6); |
|
stream_ptr += chunk_size - 6; |
|
} else { |
|
|
|
for (y_ptr = 0; y_ptr < s->frame.linesize[0] * s->avctx->height; |
|
y_ptr += s->frame.linesize[0]) { |
|
|
|
pixel_countdown = s->avctx->width; |
|
pixel_ptr = 0; |
|
while (pixel_countdown > 0) { |
|
*((signed short*)(&pixels[y_ptr + pixel_ptr])) = AV_RL16(&buf[stream_ptr+pixel_ptr]); |
|
pixel_ptr += 2; |
|
pixel_countdown--; |
|
} |
|
stream_ptr += s->avctx->width*2; |
|
} |
|
} |
|
break; |
|
|
|
case FLI_MINI: |
|
/* some sort of a thumbnail? disregard this chunk... */ |
|
stream_ptr += chunk_size - 6; |
|
break; |
|
|
|
default: |
|
av_log(avctx, AV_LOG_ERROR, "Unrecognized chunk type: %d\n", chunk_type); |
|
break; |
|
} |
|
|
|
frame_size -= chunk_size; |
|
num_chunks--; |
|
} |
|
|
|
/* by the end of the chunk, the stream ptr should equal the frame |
|
* size (minus 1, possibly); if it doesn't, issue a warning */ |
|
if ((stream_ptr != buf_size) && (stream_ptr != buf_size - 1)) |
|
av_log(avctx, AV_LOG_ERROR, "Processed FLI chunk where chunk size = %d " \ |
|
"and final chunk ptr = %d\n", buf_size, stream_ptr); |
|
|
|
|
|
*data_size=sizeof(AVFrame); |
|
*(AVFrame*)data = s->frame; |
|
|
|
return buf_size; |
|
} |
|
|
|
static int flic_decode_frame_24BPP(AVCodecContext *avctx, |
|
void *data, int *data_size, |
|
uint8_t *buf, int buf_size) |
|
{ |
|
av_log(avctx, AV_LOG_ERROR, "24Bpp FLC Unsupported due to lack of test files.\n"); |
|
return -1; |
|
} |
|
|
|
static int flic_decode_frame(AVCodecContext *avctx, |
|
void *data, int *data_size, |
|
uint8_t *buf, int buf_size) |
|
{ |
|
if (avctx->pix_fmt == PIX_FMT_PAL8) { |
|
return flic_decode_frame_8BPP(avctx, data, data_size, |
|
buf, buf_size); |
|
} |
|
else if ((avctx->pix_fmt == PIX_FMT_RGB555) || |
|
(avctx->pix_fmt == PIX_FMT_RGB565)) { |
|
return flic_decode_frame_15_16BPP(avctx, data, data_size, |
|
buf, buf_size); |
|
} |
|
else if (avctx->pix_fmt == PIX_FMT_BGR24) { |
|
return flic_decode_frame_24BPP(avctx, data, data_size, |
|
buf, buf_size); |
|
} |
|
|
|
/* Should not get here, ever as the pix_fmt is processed */ |
|
/* in flic_decode_init and the above if should deal with */ |
|
/* the finite set of possibilites allowable by here. */ |
|
/* But in case we do, just error out. */ |
|
av_log(avctx, AV_LOG_ERROR, "Unknown FLC format, my science cannot explain how this happened.\n"); |
|
return -1; |
|
} |
|
|
|
|
|
static int flic_decode_end(AVCodecContext *avctx) |
|
{ |
|
FlicDecodeContext *s = avctx->priv_data; |
|
|
|
if (s->frame.data[0]) |
|
avctx->release_buffer(avctx, &s->frame); |
|
|
|
return 0; |
|
} |
|
|
|
AVCodec flic_decoder = { |
|
"flic", |
|
CODEC_TYPE_VIDEO, |
|
CODEC_ID_FLIC, |
|
sizeof(FlicDecodeContext), |
|
flic_decode_init, |
|
NULL, |
|
flic_decode_end, |
|
flic_decode_frame, |
|
CODEC_CAP_DR1, |
|
NULL, |
|
NULL, |
|
NULL, |
|
NULL |
|
};
|
|
|