AOMedia Codec SDK
lightfield_decoder
1/*
2 * Copyright (c) 2017, Alliance for Open Media. All rights reserved
3 *
4 * This source code is subject to the terms of the BSD 2 Clause License and
5 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
6 * was not distributed with this source code in the LICENSE file, you can
7 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
8 * Media Patent License 1.0 was not distributed with this source code in the
9 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
10 */
11
12// Lightfield Decoder
13// ==================
14//
15// This is an example of a simple lightfield decoder. It builds upon the
16// simple_decoder.c example. It takes an input file containing the compressed
17// data (in ivf format), treating it as a lightfield instead of a video; and a
18// text file with a list of tiles to decode. There is an optional parameter
19// allowing to choose the output format, and the supported formats are
20// YUV1D(default), YUV, and NV12.
21// After running the lightfield encoder, run lightfield decoder to decode a
22// batch of tiles:
23// examples/lightfield_decoder vase10x10.ivf vase_reference.yuv 4 tile_list.txt
24// 0(optional)
25// The tile_list.txt is expected to be of the form:
26// Frame <frame_index0>
27// <image_index0> <anchor_index0> <tile_col0> <tile_row0>
28// <image_index1> <anchor_index1> <tile_col1> <tile_row1>
29// ...
30// Frame <frame_index1)
31// ...
32//
33// The "Frame" markers indicate a new render frame and thus a new tile list
34// will be started and the old one flushed. The image_indexN, anchor_indexN,
35// tile_colN, and tile_rowN identify an individual tile to be decoded and
36// to use anchor_indexN anchor image for MCP.
37
38#include <stdio.h>
39#include <stdlib.h>
40#include <string.h>
41
42#include "aom/aom_decoder.h"
43#include "aom/aomdx.h"
44#include "aom_scale/yv12config.h"
45#include "av1/common/enums.h"
46#include "common/tools_common.h"
47#include "common/video_reader.h"
48
49static const char *exec_name;
50
51void usage_exit(void) {
52 fprintf(stderr,
53 "Usage: %s <infile> <outfile> <num_references> <tile_list> <output "
54 "format(optional)>\n",
55 exec_name);
56 exit(EXIT_FAILURE);
57}
58
59// Output frame size
60const int output_frame_width = 512;
61const int output_frame_height = 512;
62
63static void aom_img_copy_tile(const aom_image_t *src, const aom_image_t *dst,
64 int dst_row_offset, int dst_col_offset) {
65 const int shift = (src->fmt & AOM_IMG_FMT_HIGHBITDEPTH) ? 1 : 0;
66 int plane;
67
68 for (plane = 0; plane < 3; ++plane) {
69 const unsigned char *src_buf = src->planes[plane];
70 const int src_stride = src->stride[plane];
71 unsigned char *dst_buf = dst->planes[plane];
72 const int dst_stride = dst->stride[plane];
73 const int roffset =
74 (plane > 0) ? dst_row_offset >> dst->y_chroma_shift : dst_row_offset;
75 const int coffset =
76 (plane > 0) ? dst_col_offset >> dst->x_chroma_shift : dst_col_offset;
77
78 // col offset needs to be adjusted for HBD.
79 dst_buf += roffset * dst_stride + (coffset << shift);
80
81 const int w = (aom_img_plane_width(src, plane) << shift);
82 const int h = aom_img_plane_height(src, plane);
83 int y;
84
85 for (y = 0; y < h; ++y) {
86 memcpy(dst_buf, src_buf, w);
87 src_buf += src_stride;
88 dst_buf += dst_stride;
89 }
90 }
91}
92
93void decode_tile(aom_codec_ctx_t *codec, const unsigned char *frame,
94 size_t frame_size, int tr, int tc, int ref_idx,
95 aom_image_t *reference_images, aom_image_t *output,
96 int *tile_idx, unsigned int *output_bit_depth,
97 aom_image_t **img_ptr, int output_format) {
101 aom_codec_control_(codec, AV1_SET_DECODE_TILE_COL, tc);
102
103 av1_ref_frame_t ref;
104 ref.idx = 0;
105 ref.use_external_ref = 1;
106 ref.img = reference_images[ref_idx];
107 if (aom_codec_control(codec, AV1_SET_REFERENCE, &ref)) {
108 die_codec(codec, "Failed to set reference frame.");
109 }
110
111 aom_codec_err_t aom_status = aom_codec_decode(codec, frame, frame_size, NULL);
112 if (aom_status) die_codec(codec, "Failed to decode tile.");
113
114 aom_codec_iter_t iter = NULL;
115 aom_image_t *img = aom_codec_get_frame(codec, &iter);
116 if (!img) die_codec(codec, "Failed to get frame.");
117 *img_ptr = img;
118
119 // aom_img_alloc() sets bit_depth as follows:
120 // output->bit_depth = (fmt & AOM_IMG_FMT_HIGHBITDEPTH) ? 16 : 8;
121 // Use img->bit_depth(read from bitstream), so that aom_shift_img()
122 // works as expected.
123 output->bit_depth = img->bit_depth;
124 *output_bit_depth = img->bit_depth;
125
126 if (output_format != YUV1D) {
127 // read out the tile size.
128 unsigned int tile_size = 0;
129 if (aom_codec_control(codec, AV1D_GET_TILE_SIZE, &tile_size))
130 die_codec(codec, "Failed to get the tile size");
131 const unsigned int tile_width = tile_size >> 16;
132 const unsigned int tile_height = tile_size & 65535;
133 const uint32_t output_frame_width_in_tiles =
134 output_frame_width / tile_width;
135
136 // Copy the tile to the output frame.
137 const int row_offset =
138 (*tile_idx / output_frame_width_in_tiles) * tile_height;
139 const int col_offset =
140 (*tile_idx % output_frame_width_in_tiles) * tile_width;
141
142 aom_img_copy_tile(img, output, row_offset, col_offset);
143 (*tile_idx)++;
144 }
145}
146
147static void img_write_to_file(const aom_image_t *img, FILE *file,
148 int output_format) {
149 if (output_format == YUV)
150 aom_img_write(img, file);
151 else if (output_format == NV12)
152 aom_img_write_nv12(img, file);
153 else
154 die("Invalid output format");
155}
156
157int main(int argc, char **argv) {
158 FILE *outfile = NULL;
159 aom_codec_ctx_t codec;
160 AvxVideoReader *reader = NULL;
161 const AvxInterface *decoder = NULL;
162 const AvxVideoInfo *info = NULL;
163 int num_references;
164 aom_img_fmt_t ref_fmt = 0;
165 aom_image_t reference_images[MAX_EXTERNAL_REFERENCES];
166 aom_image_t output;
167 aom_image_t *output_shifted = NULL;
168 size_t frame_size = 0;
169 const unsigned char *frame = NULL;
170 int i, j;
171 const char *tile_list_file = NULL;
172 int output_format = YUV1D;
173 exec_name = argv[0];
174
175 if (argc < 5) die("Invalid number of arguments.");
176
177 reader = aom_video_reader_open(argv[1]);
178 if (!reader) die("Failed to open %s for reading.", argv[1]);
179
180 if (!(outfile = fopen(argv[2], "wb")))
181 die("Failed to open %s for writing.", argv[2]);
182
183 num_references = (int)strtol(argv[3], NULL, 0);
184 tile_list_file = argv[4];
185
186 if (argc > 5) output_format = (int)strtol(argv[5], NULL, 0);
187 if (output_format < YUV1D || output_format > NV12)
188 die("Output format out of range [0, 2]");
189
190 info = aom_video_reader_get_info(reader);
191
192 if (info->codec_fourcc == LST_FOURCC)
193 decoder = get_aom_decoder_by_fourcc(AV1_FOURCC);
194 else
195 die("Unknown input codec.");
196 printf("Using %s\n", aom_codec_iface_name(decoder->codec_interface()));
197
198 if (aom_codec_dec_init(&codec, decoder->codec_interface(), NULL, 0))
199 die_codec(&codec, "Failed to initialize decoder.");
200
201 if (aom_codec_control(&codec, AV1D_SET_IS_ANNEXB, info->is_annexb)) {
202 die("Failed to set annex b status");
203 }
204
205 // Decode anchor frames.
207 for (i = 0; i < num_references; ++i) {
208 aom_video_reader_read_frame(reader);
209 frame = aom_video_reader_get_frame(reader, &frame_size);
210 if (aom_codec_decode(&codec, frame, frame_size, NULL))
211 die_codec(&codec, "Failed to decode frame.");
212
213 if (i == 0) {
214 if (aom_codec_control(&codec, AV1D_GET_IMG_FORMAT, &ref_fmt))
215 die_codec(&codec, "Failed to get the image format");
216
217 int frame_res[2];
218 if (aom_codec_control(&codec, AV1D_GET_FRAME_SIZE, frame_res))
219 die_codec(&codec, "Failed to get the image frame size");
220
221 // Allocate memory to store decoded references. Allocate memory with the
222 // border so that it can be used as a reference.
223 for (j = 0; j < num_references; j++) {
224 unsigned int border = AOM_DEC_BORDER_IN_PIXELS;
225 if (!aom_img_alloc_with_border(&reference_images[j], ref_fmt,
226 frame_res[0], frame_res[1], 32, 8,
227 border)) {
228 die("Failed to allocate references.");
229 }
230 }
231 }
232
234 &reference_images[i]))
235 die_codec(&codec, "Failed to copy decoded reference frame");
236
237 aom_codec_iter_t iter = NULL;
238 aom_image_t *img = NULL;
239 while ((img = aom_codec_get_frame(&codec, &iter)) != NULL) {
240 char name[1024];
241 snprintf(name, sizeof(name), "ref_%d.yuv", i);
242 printf("writing ref image to %s, %d, %d\n", name, img->d_w, img->d_h);
243 FILE *ref_file = fopen(name, "wb");
244 aom_img_write(img, ref_file);
245 fclose(ref_file);
246 }
247 }
248
249 FILE *infile = aom_video_reader_get_file(reader);
250 // Record the offset of the first camera image.
251 const FileOffset camera_frame_pos = ftello(infile);
252
253 printf("Loading compressed frames into memory.\n");
254
255 // Count the frames in the lightfield.
256 int num_frames = 0;
257 while (aom_video_reader_read_frame(reader)) {
258 ++num_frames;
259 }
260 if (num_frames < 1) die("Input light field has no frames.");
261
262 // Read all of the lightfield frames into memory.
263 unsigned char **frames =
264 (unsigned char **)malloc(num_frames * sizeof(unsigned char *));
265 size_t *frame_sizes = (size_t *)malloc(num_frames * sizeof(size_t));
266 // Seek to the first camera image.
267 fseeko(infile, camera_frame_pos, SEEK_SET);
268 for (int f = 0; f < num_frames; ++f) {
269 aom_video_reader_read_frame(reader);
270 frame = aom_video_reader_get_frame(reader, &frame_size);
271 frames[f] = (unsigned char *)malloc(frame_size * sizeof(unsigned char));
272 memcpy(frames[f], frame, frame_size);
273 frame_sizes[f] = frame_size;
274 }
275 printf("Read %d frames.\n", num_frames);
276
277 if (output_format != YUV1D) {
278 // Allocate the output frame.
279 aom_img_fmt_t out_fmt = ref_fmt;
280 if (!CONFIG_LOWBITDEPTH) out_fmt |= AOM_IMG_FMT_HIGHBITDEPTH;
281 if (!aom_img_alloc(&output, out_fmt, output_frame_width,
282 output_frame_height, 32))
283 die("Failed to allocate output image.");
284 }
285
286 printf("Decoding tile list from file.\n");
287 char line[1024];
288 FILE *tile_list_fptr = fopen(tile_list_file, "r");
289 int tile_list_cnt = 0;
290 int tile_list_writes = 0;
291 int tile_idx = 0;
292 aom_image_t *out = NULL;
293 unsigned int output_bit_depth = 0;
294
295 while ((fgets(line, 1024, tile_list_fptr)) != NULL) {
296 if (line[0] == 'F') {
297 if (output_format != YUV1D) {
298 // Write out the tile list.
299 if (tile_list_cnt) {
300 out = &output;
301 if (output_bit_depth != 0)
302 aom_shift_img(output_bit_depth, &out, &output_shifted);
303 img_write_to_file(out, outfile, output_format);
304 tile_list_writes++;
305 }
306
307 tile_list_cnt++;
308 tile_idx = 0;
309 // Then memset the frame.
310 memset(output.img_data, 0, output.sz);
311 }
312 continue;
313 }
314
315 int image_idx, ref_idx, tc, tr;
316 sscanf(line, "%d %d %d %d", &image_idx, &ref_idx, &tc, &tr);
317 if (image_idx >= num_frames) {
318 die("Tile list image_idx out of bounds: %d >= %d.", image_idx,
319 num_frames);
320 }
321 if (ref_idx >= num_references) {
322 die("Tile list ref_idx out of bounds: %d >= %d.", ref_idx,
323 num_references);
324 }
325 frame = frames[image_idx];
326 frame_size = frame_sizes[image_idx];
327
328 aom_image_t *img = NULL;
329 decode_tile(&codec, frame, frame_size, tr, tc, ref_idx, reference_images,
330 &output, &tile_idx, &output_bit_depth, &img, output_format);
331 if (output_format == YUV1D) {
332 out = img;
333 if (output_bit_depth != 0)
334 aom_shift_img(output_bit_depth, &out, &output_shifted);
335 aom_img_write(out, outfile);
336 }
337 }
338
339 if (output_format != YUV1D) {
340 // Write out the last tile list.
341 if (tile_list_writes < tile_list_cnt) {
342 out = &output;
343 if (output_bit_depth != 0)
344 aom_shift_img(output_bit_depth, &out, &output_shifted);
345 img_write_to_file(out, outfile, output_format);
346 }
347 }
348
349 if (output_shifted) aom_img_free(output_shifted);
350 if (output_format != YUV1D) aom_img_free(&output);
351 for (i = 0; i < num_references; i++) aom_img_free(&reference_images[i]);
352 for (int f = 0; f < num_frames; ++f) {
353 free(frames[f]);
354 }
355 free(frame_sizes);
356 free(frames);
357 if (aom_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec");
358 aom_video_reader_close(reader);
359 fclose(outfile);
360
361 return EXIT_SUCCESS;
362}
Describes the decoder algorithm interface to applications.
aom_image_t * aom_img_alloc_with_border(aom_image_t *img, aom_img_fmt_t fmt, unsigned int d_w, unsigned int d_h, unsigned int align, unsigned int size_align, unsigned int border)
Open a descriptor, allocating storage for the underlying image with a border.
#define AOM_IMG_FMT_HIGHBITDEPTH
Definition: aom_image.h:38
aom_image_t * aom_img_alloc(aom_image_t *img, aom_img_fmt_t fmt, unsigned int d_w, unsigned int d_h, unsigned int align)
Open a descriptor, allocating storage for the underlying image.
enum aom_img_fmt aom_img_fmt_t
List of supported image formats.
int aom_img_plane_height(const aom_image_t *img, int plane)
Get the height of a plane.
int aom_img_plane_width(const aom_image_t *img, int plane)
Get the width of a plane.
void aom_img_free(aom_image_t *img)
Close an image descriptor.
Provides definitions for using AOM or AV1 within the aom Decoder interface.
@ AV1_SET_TILE_MODE
Definition: aomdx.h:196
@ AV1D_GET_FRAME_SIZE
Definition: aomdx.h:128
@ AV1D_SET_IS_ANNEXB
Definition: aomdx.h:222
@ AV1D_GET_TILE_SIZE
Definition: aomdx.h:142
@ AV1_SET_DECODE_TILE_ROW
Definition: aomdx.h:190
@ AV1D_GET_IMG_FORMAT
Definition: aomdx.h:139
@ AV1D_EXT_TILE_DEBUG
Definition: aomdx.h:213
@ AV1_SET_REFERENCE
Definition: aom.h:60
@ AV1_COPY_NEW_FRAME_IMAGE
Definition: aom.h:66
const char * aom_codec_iface_name(aom_codec_iface_t *iface)
Return the name for a given interface.
aom_codec_err_t aom_codec_control_(aom_codec_ctx_t *ctx, int ctrl_id,...)
Control algorithm.
#define aom_codec_control(ctx, id, data)
aom_codec_control wrapper macro
Definition: aom_codec.h:414
aom_codec_err_t aom_codec_destroy(aom_codec_ctx_t *ctx)
Destroy a codec instance.
aom_codec_err_t
Algorithm return codes.
Definition: aom_codec.h:101
const void * aom_codec_iter_t
Iterator.
Definition: aom_codec.h:194
aom_image_t * aom_codec_get_frame(aom_codec_ctx_t *ctx, aom_codec_iter_t *iter)
Decoded frames iterator.
aom_codec_err_t aom_codec_decode(aom_codec_ctx_t *ctx, const uint8_t *data, size_t data_sz, void *user_priv)
Decode data.
#define aom_codec_dec_init(ctx, iface, cfg, flags)
Convenience macro for aom_codec_dec_init_ver()
Definition: aom_decoder.h:135
Codec context structure.
Definition: aom_codec.h:204
Image Descriptor.
Definition: aom_image.h:141
unsigned int bit_depth
Definition: aom_image.h:153
unsigned int y_chroma_shift
Definition: aom_image.h:165
aom_img_fmt_t fmt
Definition: aom_image.h:142
int stride[3]
Definition: aom_image.h:173
unsigned char * img_data
Definition: aom_image.h:187
unsigned int x_chroma_shift
Definition: aom_image.h:164
unsigned int d_w
Definition: aom_image.h:156
unsigned int d_h
Definition: aom_image.h:157
unsigned char * planes[3]
Definition: aom_image.h:172
size_t sz
Definition: aom_image.h:174
AV1 specific reference frame data struct.
Definition: aom.h:108
int use_external_ref
Definition: aom.h:110
aom_image_t img
Definition: aom.h:111
int idx
Definition: aom.h:109