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
49 static const char *exec_name;
50
usage_exit(void)51 void 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
60 const int output_frame_width = 512;
61 const int output_frame_height = 512;
62
aom_img_copy_tile(const aom_image_t * src,const aom_image_t * dst,int dst_row_offset,int dst_col_offset)63 static 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
decode_tile(aom_codec_ctx_t * codec,const unsigned char * frame,size_t frame_size,int tr,int tc,int ref_idx,aom_image_t * reference_images,aom_image_t * output,int * tile_idx,unsigned int * output_bit_depth,aom_image_t ** img_ptr,int output_format)93 void 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) {
98 AOM_CODEC_CONTROL_TYPECHECKED(codec, AV1_SET_TILE_MODE, 1);
99 AOM_CODEC_CONTROL_TYPECHECKED(codec, AV1D_EXT_TILE_DEBUG, 1);
100 AOM_CODEC_CONTROL_TYPECHECKED(codec, AV1_SET_DECODE_TILE_ROW, tr);
101 AOM_CODEC_CONTROL_TYPECHECKED(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_TYPECHECKED(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_TYPECHECKED(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
img_write_to_file(const aom_image_t * img,FILE * file,int output_format)147 static 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
main(int argc,char ** argv)157 int main(int argc, char **argv) {
158 FILE *outfile = NULL;
159 AvxVideoReader *reader = NULL;
160 const AvxVideoInfo *info = NULL;
161 int num_references;
162 aom_img_fmt_t ref_fmt = 0;
163 aom_image_t reference_images[MAX_EXTERNAL_REFERENCES];
164 aom_image_t output;
165 aom_image_t *output_shifted = NULL;
166 size_t frame_size = 0;
167 const unsigned char *frame = NULL;
168 int i, j;
169 const char *tile_list_file = NULL;
170 int output_format = YUV1D;
171 exec_name = argv[0];
172
173 if (argc < 5) die("Invalid number of arguments.");
174
175 reader = aom_video_reader_open(argv[1]);
176 if (!reader) die("Failed to open %s for reading.", argv[1]);
177
178 if (!(outfile = fopen(argv[2], "wb")))
179 die("Failed to open %s for writing.", argv[2]);
180
181 num_references = (int)strtol(argv[3], NULL, 0);
182 tile_list_file = argv[4];
183
184 if (argc > 5) output_format = (int)strtol(argv[5], NULL, 0);
185 if (output_format < YUV1D || output_format > NV12)
186 die("Output format out of range [0, 2]");
187
188 info = aom_video_reader_get_info(reader);
189
190 aom_codec_iface_t *decoder;
191 if (info->codec_fourcc == LST_FOURCC)
192 decoder = get_aom_decoder_by_fourcc(AV1_FOURCC);
193 else
194 die("Unknown input codec.");
195 printf("Using %s\n", aom_codec_iface_name(decoder));
196
197 aom_codec_ctx_t codec;
198 if (aom_codec_dec_init(&codec, decoder, NULL, 0))
199 die_codec(&codec, "Failed to initialize decoder.");
200
201 if (AOM_CODEC_CONTROL_TYPECHECKED(&codec, AV1D_SET_IS_ANNEXB,
202 info->is_annexb)) {
203 die("Failed to set annex b status");
204 }
205
206 // Decode anchor frames.
207 AOM_CODEC_CONTROL_TYPECHECKED(&codec, AV1_SET_TILE_MODE, 0);
208 for (i = 0; i < num_references; ++i) {
209 aom_video_reader_read_frame(reader);
210 frame = aom_video_reader_get_frame(reader, &frame_size);
211 if (aom_codec_decode(&codec, frame, frame_size, NULL))
212 die_codec(&codec, "Failed to decode frame.");
213
214 if (i == 0) {
215 if (AOM_CODEC_CONTROL_TYPECHECKED(&codec, AV1D_GET_IMG_FORMAT, &ref_fmt))
216 die_codec(&codec, "Failed to get the image format");
217
218 int frame_res[2];
219 if (AOM_CODEC_CONTROL_TYPECHECKED(&codec, AV1D_GET_FRAME_SIZE, frame_res))
220 die_codec(&codec, "Failed to get the image frame size");
221
222 // Allocate memory to store decoded references. Allocate memory with the
223 // border so that it can be used as a reference.
224 for (j = 0; j < num_references; j++) {
225 unsigned int border = AOM_DEC_BORDER_IN_PIXELS;
226 if (!aom_img_alloc_with_border(&reference_images[j], ref_fmt,
227 frame_res[0], frame_res[1], 32, 8,
228 border)) {
229 die("Failed to allocate references.");
230 }
231 }
232 }
233
234 if (AOM_CODEC_CONTROL_TYPECHECKED(&codec, AV1_COPY_NEW_FRAME_IMAGE,
235 &reference_images[i]))
236 die_codec(&codec, "Failed to copy decoded reference frame");
237
238 aom_codec_iter_t iter = NULL;
239 aom_image_t *img = NULL;
240 while ((img = aom_codec_get_frame(&codec, &iter)) != NULL) {
241 char name[1024];
242 snprintf(name, sizeof(name), "ref_%d.yuv", i);
243 printf("writing ref image to %s, %u, %u\n", name, img->d_w, img->d_h);
244 FILE *ref_file = fopen(name, "wb");
245 aom_img_write(img, ref_file);
246 fclose(ref_file);
247 }
248 }
249
250 FILE *infile = aom_video_reader_get_file(reader);
251 // Record the offset of the first camera image.
252 const FileOffset camera_frame_pos = ftello(infile);
253
254 printf("Loading compressed frames into memory.\n");
255
256 // Count the frames in the lightfield.
257 int num_frames = 0;
258 while (aom_video_reader_read_frame(reader)) {
259 ++num_frames;
260 }
261 if (num_frames < 1) die("Input light field has no frames.");
262
263 // Read all of the lightfield frames into memory.
264 unsigned char **frames =
265 (unsigned char **)malloc(num_frames * sizeof(unsigned char *));
266 size_t *frame_sizes = (size_t *)malloc(num_frames * sizeof(size_t));
267 // Seek to the first camera image.
268 fseeko(infile, camera_frame_pos, SEEK_SET);
269 for (int f = 0; f < num_frames; ++f) {
270 aom_video_reader_read_frame(reader);
271 frame = aom_video_reader_get_frame(reader, &frame_size);
272 frames[f] = (unsigned char *)malloc(frame_size * sizeof(unsigned char));
273 memcpy(frames[f], frame, frame_size);
274 frame_sizes[f] = frame_size;
275 }
276 printf("Read %d frames.\n", num_frames);
277
278 if (output_format != YUV1D) {
279 // Allocate the output frame.
280 aom_img_fmt_t out_fmt = ref_fmt;
281 if (FORCE_HIGHBITDEPTH_DECODING) out_fmt |= AOM_IMG_FMT_HIGHBITDEPTH;
282 if (!aom_img_alloc(&output, out_fmt, output_frame_width,
283 output_frame_height, 32))
284 die("Failed to allocate output image.");
285 }
286
287 printf("Decoding tile list from file.\n");
288 char line[1024];
289 FILE *tile_list_fptr = fopen(tile_list_file, "r");
290 if (!tile_list_fptr) die_codec(&codec, "Failed to open tile list file.");
291 int tile_list_cnt = 0;
292 int tile_list_writes = 0;
293 int tile_idx = 0;
294 aom_image_t *out = NULL;
295 unsigned int output_bit_depth = 0;
296
297 while ((fgets(line, 1024, tile_list_fptr)) != NULL) {
298 if (line[0] == 'F') {
299 if (output_format != YUV1D) {
300 // Write out the tile list.
301 if (tile_list_cnt) {
302 out = &output;
303 if (output_bit_depth != 0)
304 aom_shift_img(output_bit_depth, &out, &output_shifted);
305 img_write_to_file(out, outfile, output_format);
306 tile_list_writes++;
307 }
308
309 tile_list_cnt++;
310 tile_idx = 0;
311 // Then memset the frame.
312 memset(output.img_data, 0, output.sz);
313 }
314 continue;
315 }
316
317 int image_idx, ref_idx, tc, tr;
318 sscanf(line, "%d %d %d %d", &image_idx, &ref_idx, &tc, &tr);
319 if (image_idx >= num_frames) {
320 die("Tile list image_idx out of bounds: %d >= %d.", image_idx,
321 num_frames);
322 }
323 if (ref_idx >= num_references) {
324 die("Tile list ref_idx out of bounds: %d >= %d.", ref_idx,
325 num_references);
326 }
327 frame = frames[image_idx];
328 frame_size = frame_sizes[image_idx];
329
330 aom_image_t *img = NULL;
331 decode_tile(&codec, frame, frame_size, tr, tc, ref_idx, reference_images,
332 &output, &tile_idx, &output_bit_depth, &img, output_format);
333 if (output_format == YUV1D) {
334 out = img;
335 if (output_bit_depth != 0)
336 aom_shift_img(output_bit_depth, &out, &output_shifted);
337 aom_img_write(out, outfile);
338 }
339 }
340
341 if (output_format != YUV1D) {
342 // Write out the last tile list.
343 if (tile_list_writes < tile_list_cnt) {
344 out = &output;
345 if (output_bit_depth != 0)
346 aom_shift_img(output_bit_depth, &out, &output_shifted);
347 img_write_to_file(out, outfile, output_format);
348 }
349 }
350
351 if (output_shifted) aom_img_free(output_shifted);
352 if (output_format != YUV1D) aom_img_free(&output);
353 for (i = 0; i < num_references; i++) aom_img_free(&reference_images[i]);
354 for (int f = 0; f < num_frames; ++f) {
355 free(frames[f]);
356 }
357 free(frame_sizes);
358 free(frames);
359 if (aom_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec");
360 aom_video_reader_close(reader);
361 fclose(outfile);
362
363 return EXIT_SUCCESS;
364 }
365