xref: /third_party/ffmpeg/libavcodec/rawdec.c (revision cabdff1a)
1/*
2 * Raw Video Decoder
3 * Copyright (c) 2001 Fabrice Bellard
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22/**
23 * @file
24 * Raw Video Decoder
25 */
26
27#include "avcodec.h"
28#include "bswapdsp.h"
29#include "codec_internal.h"
30#include "decode.h"
31#include "get_bits.h"
32#include "internal.h"
33#include "raw.h"
34#include "libavutil/avassert.h"
35#include "libavutil/buffer.h"
36#include "libavutil/common.h"
37#include "libavutil/intreadwrite.h"
38#include "libavutil/imgutils.h"
39#include "libavutil/opt.h"
40
41typedef struct RawVideoContext {
42    AVClass *av_class;
43    AVBufferRef *palette;
44    int frame_size;  /* size of the frame in bytes */
45    int flip;
46    int is_1_2_4_8_bpp; // 1, 2, 4 and 8 bpp in avi/mov, 1 and 8 bpp in nut
47    int is_mono;
48    int is_pal8;
49    int is_nut_mono;
50    int is_nut_pal8;
51    int is_yuv2;
52    int is_lt_16bpp; // 16bpp pixfmt and bits_per_coded_sample < 16
53    int tff;
54
55    BswapDSPContext bbdsp;
56    void *bitstream_buf;
57    unsigned int bitstream_buf_size;
58} RawVideoContext;
59
60static const AVOption options[]={
61{"top", "top field first", offsetof(RawVideoContext, tff), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, AV_OPT_FLAG_DECODING_PARAM|AV_OPT_FLAG_VIDEO_PARAM},
62{NULL}
63};
64
65static const AVClass rawdec_class = {
66    .class_name = "rawdec",
67    .option     = options,
68    .version    = LIBAVUTIL_VERSION_INT,
69};
70
71static av_cold int raw_init_decoder(AVCodecContext *avctx)
72{
73    RawVideoContext *context = avctx->priv_data;
74    const AVPixFmtDescriptor *desc;
75
76    ff_bswapdsp_init(&context->bbdsp);
77
78    if (   avctx->codec_tag == MKTAG('r','a','w',' ')
79        || avctx->codec_tag == MKTAG('N','O','1','6'))
80        avctx->pix_fmt = avpriv_pix_fmt_find(PIX_FMT_LIST_MOV,
81                                      avctx->bits_per_coded_sample);
82    else if (avctx->codec_tag == MKTAG('W', 'R', 'A', 'W'))
83        avctx->pix_fmt = avpriv_pix_fmt_find(PIX_FMT_LIST_AVI,
84                                      avctx->bits_per_coded_sample);
85    else if (avctx->codec_tag && (avctx->codec_tag & 0xFFFFFF) != MKTAG('B','I','T', 0))
86        avctx->pix_fmt = avpriv_pix_fmt_find(PIX_FMT_LIST_RAW, avctx->codec_tag);
87    else if (avctx->pix_fmt == AV_PIX_FMT_NONE && avctx->bits_per_coded_sample)
88        avctx->pix_fmt = avpriv_pix_fmt_find(PIX_FMT_LIST_AVI,
89                                      avctx->bits_per_coded_sample);
90
91    desc = av_pix_fmt_desc_get(avctx->pix_fmt);
92    if (!desc) {
93        av_log(avctx, AV_LOG_ERROR, "Invalid pixel format.\n");
94        return AVERROR(EINVAL);
95    }
96
97    if (desc->flags & AV_PIX_FMT_FLAG_PAL) {
98        context->palette = av_buffer_alloc(AVPALETTE_SIZE);
99        if (!context->palette)
100            return AVERROR(ENOMEM);
101        memset(context->palette->data, 0, AVPALETTE_SIZE);
102        if (avctx->bits_per_coded_sample == 1)
103            memset(context->palette->data, 0xff, 4);
104    }
105
106    if ((avctx->extradata_size >= 9 &&
107         !memcmp(avctx->extradata + avctx->extradata_size - 9, "BottomUp", 9)) ||
108        avctx->codec_tag == MKTAG('c','y','u','v') ||
109        avctx->codec_tag == MKTAG(3, 0, 0, 0) ||
110        avctx->codec_tag == MKTAG('W','R','A','W'))
111        context->flip = 1;
112
113    if (avctx->pix_fmt == AV_PIX_FMT_MONOWHITE ||
114        avctx->pix_fmt == AV_PIX_FMT_MONOBLACK)
115        context->is_mono = 1;
116    else if (avctx->pix_fmt == AV_PIX_FMT_PAL8)
117        context->is_pal8 = 1;
118
119    if (avctx->codec_tag == MKTAG('B','1','W','0') ||
120        avctx->codec_tag == MKTAG('B','0','W','1'))
121        context->is_nut_mono = 1;
122    else if (avctx->codec_tag == MKTAG('P','A','L',8))
123        context->is_nut_pal8 = 1;
124
125    if (avctx->codec_tag == AV_RL32("yuv2") &&
126        avctx->pix_fmt   == AV_PIX_FMT_YUYV422)
127        context->is_yuv2 = 1;
128
129    return 0;
130}
131
132static void flip(AVCodecContext *avctx, AVFrame *frame)
133{
134    frame->data[0]     += frame->linesize[0] * (avctx->height - 1);
135    frame->linesize[0] *= -1;
136}
137
138/*
139 * Scale sample to 16-bit resolution
140 */
141#define SCALE16(x, bits) (((x) << (16 - (bits))) | ((x) >> (2 * (bits) - 16)))
142
143/**
144 * Scale buffer to 16 bits per coded sample resolution
145 */
146#define MKSCALE16(name, r16, w16) \
147static void name(AVCodecContext *avctx, uint8_t * dst, const uint8_t *buf, int buf_size, int packed) \
148{ \
149    int i; \
150    if (!packed) { \
151        for (i = 0; i + 1 < buf_size; i += 2) \
152            w16(dst + i, SCALE16(r16(buf + i), avctx->bits_per_coded_sample)); \
153    } else { \
154        GetBitContext gb; \
155        init_get_bits(&gb, buf, buf_size * 8); \
156        for (i = 0; i < avctx->width * avctx->height; i++) { \
157            int sample = get_bits(&gb, avctx->bits_per_coded_sample); \
158            w16(dst + i*2, SCALE16(sample, avctx->bits_per_coded_sample)); \
159        } \
160   } \
161}
162
163MKSCALE16(scale16be, AV_RB16, AV_WB16)
164MKSCALE16(scale16le, AV_RL16, AV_WL16)
165
166static int raw_decode(AVCodecContext *avctx, AVFrame *frame,
167                      int *got_frame, AVPacket *avpkt)
168{
169    const AVPixFmtDescriptor *desc;
170    RawVideoContext *context       = avctx->priv_data;
171    const uint8_t *buf             = avpkt->data;
172    int buf_size                   = avpkt->size;
173    int linesize_align             = 4;
174    int stride;
175    int res, len;
176    int need_copy;
177
178    if (avctx->width <= 0) {
179        av_log(avctx, AV_LOG_ERROR, "width is not set\n");
180        return AVERROR_INVALIDDATA;
181    }
182    if (avctx->height <= 0) {
183        av_log(avctx, AV_LOG_ERROR, "height is not set\n");
184        return AVERROR_INVALIDDATA;
185    }
186
187    if (context->is_nut_mono)
188        stride = avctx->width / 8 + (avctx->width & 7 ? 1 : 0);
189    else if (context->is_nut_pal8)
190        stride = avctx->width;
191    else
192        stride = avpkt->size / avctx->height;
193
194    av_log(avctx, AV_LOG_DEBUG, "PACKET SIZE: %d, STRIDE: %d\n", avpkt->size, stride);
195
196    if (stride == 0 || avpkt->size < stride * avctx->height) {
197        av_log(avctx, AV_LOG_ERROR, "Packet too small (%d)\n", avpkt->size);
198        return AVERROR_INVALIDDATA;
199    }
200
201    desc = av_pix_fmt_desc_get(avctx->pix_fmt);
202
203    if ((avctx->bits_per_coded_sample == 8 || avctx->bits_per_coded_sample == 4 ||
204         avctx->bits_per_coded_sample == 2 || avctx->bits_per_coded_sample == 1 ||
205         (avctx->bits_per_coded_sample == 0 && (context->is_nut_pal8 || context->is_mono)) ) &&
206        (context->is_mono || context->is_pal8) &&
207        (!avctx->codec_tag || avctx->codec_tag == MKTAG('r','a','w',' ') ||
208                context->is_nut_mono || context->is_nut_pal8)) {
209        context->is_1_2_4_8_bpp = 1;
210        if (context->is_mono) {
211            int row_bytes = avctx->width / 8 + (avctx->width & 7 ? 1 : 0);
212            context->frame_size = av_image_get_buffer_size(avctx->pix_fmt,
213                                                           FFALIGN(row_bytes, 16) * 8,
214                                                           avctx->height, 1);
215        } else
216            context->frame_size = av_image_get_buffer_size(avctx->pix_fmt,
217                                                           FFALIGN(avctx->width, 16),
218                                                           avctx->height, 1);
219    } else {
220        context->is_lt_16bpp = av_get_bits_per_pixel(desc) == 16 && avctx->bits_per_coded_sample > 8 && avctx->bits_per_coded_sample < 16;
221        context->frame_size = av_image_get_buffer_size(avctx->pix_fmt, avctx->width,
222                                                       avctx->height, 1);
223    }
224    if (context->frame_size < 0)
225        return context->frame_size;
226
227    need_copy = !avpkt->buf || context->is_1_2_4_8_bpp || context->is_yuv2 || context->is_lt_16bpp;
228
229    frame->pict_type        = AV_PICTURE_TYPE_I;
230    frame->key_frame        = 1;
231
232    res = ff_decode_frame_props(avctx, frame);
233    if (res < 0)
234        return res;
235
236    frame->pkt_pos      = avctx->internal->last_pkt_props->pos;
237    frame->pkt_duration = avctx->internal->last_pkt_props->duration;
238
239    if (context->tff >= 0) {
240        frame->interlaced_frame = 1;
241        frame->top_field_first  = context->tff;
242    }
243
244    if ((res = av_image_check_size(avctx->width, avctx->height, 0, avctx)) < 0)
245        return res;
246
247    if (need_copy)
248        frame->buf[0] = av_buffer_alloc(FFMAX(context->frame_size, buf_size));
249    else
250        frame->buf[0] = av_buffer_ref(avpkt->buf);
251    if (!frame->buf[0])
252        return AVERROR(ENOMEM);
253
254    // 1, 2, 4 and 8 bpp in avi/mov, 1 and 8 bpp in nut
255    if (context->is_1_2_4_8_bpp) {
256        int i, j, row_pix = 0;
257        uint8_t *dst = frame->buf[0]->data;
258        buf_size = context->frame_size - (context->is_pal8 ? AVPALETTE_SIZE : 0);
259        if (avctx->bits_per_coded_sample == 8 || context->is_nut_pal8 || context->is_mono) {
260            int pix_per_byte = context->is_mono ? 8 : 1;
261            for (i = 0, j = 0; j < buf_size && i<avpkt->size; i++, j++) {
262                dst[j] = buf[i];
263                row_pix += pix_per_byte;
264                if (row_pix >= avctx->width) {
265                    i += stride - (i % stride) - 1;
266                    j += 16 - (j % 16) - 1;
267                    row_pix = 0;
268                }
269            }
270        } else if (avctx->bits_per_coded_sample == 4) {
271            for (i = 0, j = 0; 2 * j + 1 < buf_size && i<avpkt->size; i++, j++) {
272                dst[2 * j + 0] = buf[i] >> 4;
273                dst[2 * j + 1] = buf[i] & 15;
274                row_pix += 2;
275                if (row_pix >= avctx->width) {
276                    i += stride - (i % stride) - 1;
277                    j += 8 - (j % 8) - 1;
278                    row_pix = 0;
279                }
280            }
281        } else if (avctx->bits_per_coded_sample == 2) {
282            for (i = 0, j = 0; 4 * j + 3 < buf_size && i<avpkt->size; i++, j++) {
283                dst[4 * j + 0] = buf[i] >> 6;
284                dst[4 * j + 1] = buf[i] >> 4 & 3;
285                dst[4 * j + 2] = buf[i] >> 2 & 3;
286                dst[4 * j + 3] = buf[i]      & 3;
287                row_pix += 4;
288                if (row_pix >= avctx->width) {
289                    i += stride - (i % stride) - 1;
290                    j += 4 - (j % 4) - 1;
291                    row_pix = 0;
292                }
293            }
294        } else {
295            av_assert0(avctx->bits_per_coded_sample == 1);
296            for (i = 0, j = 0; 8 * j + 7 < buf_size && i<avpkt->size; i++, j++) {
297                dst[8 * j + 0] = buf[i] >> 7;
298                dst[8 * j + 1] = buf[i] >> 6 & 1;
299                dst[8 * j + 2] = buf[i] >> 5 & 1;
300                dst[8 * j + 3] = buf[i] >> 4 & 1;
301                dst[8 * j + 4] = buf[i] >> 3 & 1;
302                dst[8 * j + 5] = buf[i] >> 2 & 1;
303                dst[8 * j + 6] = buf[i] >> 1 & 1;
304                dst[8 * j + 7] = buf[i]      & 1;
305                row_pix += 8;
306                if (row_pix >= avctx->width) {
307                    i += stride - (i % stride) - 1;
308                    j += 2 - (j % 2) - 1;
309                    row_pix = 0;
310                }
311            }
312        }
313        linesize_align = 16;
314        buf = dst;
315    } else if (context->is_lt_16bpp) {
316        uint8_t *dst = frame->buf[0]->data;
317        int packed = (avctx->codec_tag & 0xFFFFFF) == MKTAG('B','I','T', 0);
318        int swap   =  avctx->codec_tag >> 24;
319
320        if (packed && swap) {
321            av_fast_padded_malloc(&context->bitstream_buf, &context->bitstream_buf_size, buf_size);
322            if (!context->bitstream_buf)
323                return AVERROR(ENOMEM);
324            if (swap == 16)
325                context->bbdsp.bswap16_buf(context->bitstream_buf, (const uint16_t*)buf, buf_size / 2);
326            else if (swap == 32)
327                context->bbdsp.bswap_buf(context->bitstream_buf, (const uint32_t*)buf, buf_size / 4);
328            else
329                return AVERROR_INVALIDDATA;
330            buf = context->bitstream_buf;
331        }
332
333        if (desc->flags & AV_PIX_FMT_FLAG_BE)
334            scale16be(avctx, dst, buf, buf_size, packed);
335        else
336            scale16le(avctx, dst, buf, buf_size, packed);
337
338        buf = dst;
339    } else if (need_copy) {
340        memcpy(frame->buf[0]->data, buf, buf_size);
341        buf = frame->buf[0]->data;
342    }
343
344    if (avctx->codec_tag == MKTAG('A', 'V', '1', 'x') ||
345        avctx->codec_tag == MKTAG('A', 'V', 'u', 'p'))
346        buf += buf_size - context->frame_size;
347
348    len = context->frame_size - (avctx->pix_fmt==AV_PIX_FMT_PAL8 ? AVPALETTE_SIZE : 0);
349    if (buf_size < len && ((avctx->codec_tag & 0xFFFFFF) != MKTAG('B','I','T', 0) || !need_copy)) {
350        av_log(avctx, AV_LOG_ERROR, "Invalid buffer size, packet size %d < expected frame_size %d\n", buf_size, len);
351        av_buffer_unref(&frame->buf[0]);
352        return AVERROR(EINVAL);
353    }
354
355    if ((res = av_image_fill_arrays(frame->data, frame->linesize,
356                                    buf, avctx->pix_fmt,
357                                    avctx->width, avctx->height, 1)) < 0) {
358        av_buffer_unref(&frame->buf[0]);
359        return res;
360    }
361
362    if (avctx->pix_fmt == AV_PIX_FMT_PAL8) {
363        int ret;
364
365        if (!context->palette)
366            context->palette = av_buffer_alloc(AVPALETTE_SIZE);
367        if (!context->palette) {
368            av_buffer_unref(&frame->buf[0]);
369            return AVERROR(ENOMEM);
370        }
371        ret = av_buffer_make_writable(&context->palette);
372        if (ret < 0) {
373            av_buffer_unref(&frame->buf[0]);
374            return ret;
375        }
376
377        if (ff_copy_palette(context->palette->data, avpkt, avctx)) {
378            frame->palette_has_changed = 1;
379        } else if (context->is_nut_pal8) {
380            int vid_size = avctx->width * avctx->height;
381            int pal_size = avpkt->size - vid_size;
382
383            if (avpkt->size > vid_size && pal_size <= AVPALETTE_SIZE) {
384                const uint8_t *pal = avpkt->data + vid_size;
385                memcpy(context->palette->data, pal, pal_size);
386                frame->palette_has_changed = 1;
387            }
388        }
389    }
390
391    if ((avctx->pix_fmt==AV_PIX_FMT_RGB24    ||
392        avctx->pix_fmt==AV_PIX_FMT_BGR24     ||
393        avctx->pix_fmt==AV_PIX_FMT_GRAY8     ||
394        avctx->pix_fmt==AV_PIX_FMT_RGB555LE  ||
395        avctx->pix_fmt==AV_PIX_FMT_RGB555BE  ||
396        avctx->pix_fmt==AV_PIX_FMT_RGB565LE  ||
397        avctx->pix_fmt==AV_PIX_FMT_MONOWHITE ||
398        avctx->pix_fmt==AV_PIX_FMT_MONOBLACK ||
399        avctx->pix_fmt==AV_PIX_FMT_PAL8) &&
400        FFALIGN(frame->linesize[0], linesize_align) * avctx->height <= buf_size)
401        frame->linesize[0] = FFALIGN(frame->linesize[0], linesize_align);
402
403    if (avctx->pix_fmt == AV_PIX_FMT_NV12 && avctx->codec_tag == MKTAG('N', 'V', '1', '2') &&
404        FFALIGN(frame->linesize[0], linesize_align) * avctx->height +
405        FFALIGN(frame->linesize[1], linesize_align) * ((avctx->height + 1) / 2) <= buf_size) {
406        int la0 = FFALIGN(frame->linesize[0], linesize_align);
407        frame->data[1] += (la0 - frame->linesize[0]) * avctx->height;
408        frame->linesize[0] = la0;
409        frame->linesize[1] = FFALIGN(frame->linesize[1], linesize_align);
410    }
411
412    if (avctx->pix_fmt == AV_PIX_FMT_PAL8 && buf_size < context->frame_size) {
413        frame->buf[1]  = av_buffer_ref(context->palette);
414        if (!frame->buf[1]) {
415            av_buffer_unref(&frame->buf[0]);
416            return AVERROR(ENOMEM);
417        }
418        frame->data[1] = frame->buf[1]->data;
419    }
420
421    if (avctx->pix_fmt == AV_PIX_FMT_BGR24 &&
422        ((frame->linesize[0] + 3) & ~3) * avctx->height <= buf_size)
423        frame->linesize[0] = (frame->linesize[0] + 3) & ~3;
424
425    if (context->flip)
426        flip(avctx, frame);
427
428    if (avctx->codec_tag == MKTAG('Y', 'V', '1', '2') ||
429        avctx->codec_tag == MKTAG('Y', 'V', '1', '6') ||
430        avctx->codec_tag == MKTAG('Y', 'V', '2', '4') ||
431        avctx->codec_tag == MKTAG('Y', 'V', 'U', '9'))
432        FFSWAP(uint8_t *, frame->data[1], frame->data[2]);
433
434    if (avctx->codec_tag == AV_RL32("I420") && (avctx->width+1)*(avctx->height+1) * 3/2 == buf_size) {
435        frame->data[1] = frame->data[1] +  (avctx->width+1)*(avctx->height+1) -avctx->width*avctx->height;
436        frame->data[2] = frame->data[2] + ((avctx->width+1)*(avctx->height+1) -avctx->width*avctx->height)*5/4;
437    }
438
439    if (avctx->codec_tag == AV_RL32("yuv2") &&
440        avctx->pix_fmt   == AV_PIX_FMT_YUYV422) {
441        int x, y;
442        uint8_t *line = frame->data[0];
443        for (y = 0; y < avctx->height; y++) {
444            for (x = 0; x < avctx->width; x++)
445                line[2 * x + 1] ^= 0x80;
446            line += frame->linesize[0];
447        }
448    }
449
450    if (avctx->codec_tag == AV_RL32("b64a") &&
451        avctx->pix_fmt   == AV_PIX_FMT_RGBA64BE) {
452        uint8_t *dst = frame->data[0];
453        uint64_t v;
454        int x, y;
455        for (y = 0; y < avctx->height; y++) {
456            for (x = 0; x >> 3 < avctx->width; x += 8) {
457                v = AV_RB64(&dst[x]);
458                AV_WB64(&dst[x], v << 16 | v >> 48);
459            }
460            dst += frame->linesize[0];
461        }
462    }
463
464    if (avctx->field_order > AV_FIELD_PROGRESSIVE) { /* we have interlaced material flagged in container */
465        frame->interlaced_frame = 1;
466        if (avctx->field_order == AV_FIELD_TT || avctx->field_order == AV_FIELD_TB)
467            frame->top_field_first = 1;
468    }
469
470    *got_frame = 1;
471    return buf_size;
472}
473
474static av_cold int raw_close_decoder(AVCodecContext *avctx)
475{
476    RawVideoContext *context = avctx->priv_data;
477
478    av_buffer_unref(&context->palette);
479    av_freep(&context->bitstream_buf);
480    return 0;
481}
482
483const FFCodec ff_rawvideo_decoder = {
484    .p.name         = "rawvideo",
485    .p.long_name    = NULL_IF_CONFIG_SMALL("raw video"),
486    .p.type         = AVMEDIA_TYPE_VIDEO,
487    .p.id           = AV_CODEC_ID_RAWVIDEO,
488    .priv_data_size = sizeof(RawVideoContext),
489    .init           = raw_init_decoder,
490    .close          = raw_close_decoder,
491    FF_CODEC_DECODE_CB(raw_decode),
492    .p.priv_class   = &rawdec_class,
493    .p.capabilities = AV_CODEC_CAP_PARAM_CHANGE,
494    .caps_internal  = FF_CODEC_CAP_INIT_THREADSAFE,
495};
496