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 enum {
50 YUV1D, // 1D tile output for conformance test.
51 YUV, // Tile output in YUV format.
52 NV12, // Tile output in NV12 format.
53 } UENUM1BYTE(OUTPUT_FORMAT);
54
55 static const char *exec_name;
56
usage_exit(void)57 void usage_exit(void) {
58 fprintf(stderr,
59 "Usage: %s <infile> <outfile> <num_references> <tile_list> <output "
60 "format(optional)>\n",
61 exec_name);
62 exit(EXIT_FAILURE);
63 }
64
65 // Output frame size
66 static const int output_frame_width = 512;
67 static const int output_frame_height = 512;
68
aom_img_copy_tile(const aom_image_t * src,const aom_image_t * dst,int dst_row_offset,int dst_col_offset)69 static void aom_img_copy_tile(const aom_image_t *src, const aom_image_t *dst,
70 int dst_row_offset, int dst_col_offset) {
71 const int shift = (src->fmt & AOM_IMG_FMT_HIGHBITDEPTH) ? 1 : 0;
72 int plane;
73
74 for (plane = 0; plane < 3; ++plane) {
75 const unsigned char *src_buf = src->planes[plane];
76 const int src_stride = src->stride[plane];
77 unsigned char *dst_buf = dst->planes[plane];
78 const int dst_stride = dst->stride[plane];
79 const int roffset =
80 (plane > 0) ? dst_row_offset >> dst->y_chroma_shift : dst_row_offset;
81 const int coffset =
82 (plane > 0) ? dst_col_offset >> dst->x_chroma_shift : dst_col_offset;
83
84 // col offset needs to be adjusted for HBD.
85 dst_buf += roffset * dst_stride + (coffset << shift);
86
87 const int w = (aom_img_plane_width(src, plane) << shift);
88 const int h = aom_img_plane_height(src, plane);
89 int y;
90
91 for (y = 0; y < h; ++y) {
92 memcpy(dst_buf, src_buf, w);
93 src_buf += src_stride;
94 dst_buf += dst_stride;
95 }
96 }
97 }
98
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)99 static void decode_tile(aom_codec_ctx_t *codec, const unsigned char *frame,
100 size_t frame_size, int tr, int tc, int ref_idx,
101 aom_image_t *reference_images, aom_image_t *output,
102 int *tile_idx, unsigned int *output_bit_depth,
103 aom_image_t **img_ptr, int output_format) {
104 AOM_CODEC_CONTROL_TYPECHECKED(codec, AV1_SET_TILE_MODE, 1);
105 AOM_CODEC_CONTROL_TYPECHECKED(codec, AV1D_EXT_TILE_DEBUG, 1);
106 AOM_CODEC_CONTROL_TYPECHECKED(codec, AV1_SET_DECODE_TILE_ROW, tr);
107 AOM_CODEC_CONTROL_TYPECHECKED(codec, AV1_SET_DECODE_TILE_COL, tc);
108
109 av1_ref_frame_t ref;
110 ref.idx = 0;
111 ref.use_external_ref = 1;
112 ref.img = reference_images[ref_idx];
113 if (AOM_CODEC_CONTROL_TYPECHECKED(codec, AV1_SET_REFERENCE, &ref)) {
114 die_codec(codec, "Failed to set reference frame.");
115 }
116
117 aom_codec_err_t aom_status = aom_codec_decode(codec, frame, frame_size, NULL);
118 if (aom_status) die_codec(codec, "Failed to decode tile.");
119
120 aom_codec_iter_t iter = NULL;
121 aom_image_t *img = aom_codec_get_frame(codec, &iter);
122 if (!img) die_codec(codec, "Failed to get frame.");
123 *img_ptr = img;
124
125 // aom_img_alloc() sets bit_depth as follows:
126 // output->bit_depth = (fmt & AOM_IMG_FMT_HIGHBITDEPTH) ? 16 : 8;
127 // Use img->bit_depth(read from bitstream), so that aom_shift_img()
128 // works as expected.
129 output->bit_depth = img->bit_depth;
130 *output_bit_depth = img->bit_depth;
131
132 if (output_format != YUV1D) {
133 // read out the tile size.
134 unsigned int tile_size = 0;
135 if (AOM_CODEC_CONTROL_TYPECHECKED(codec, AV1D_GET_TILE_SIZE, &tile_size))
136 die_codec(codec, "Failed to get the tile size");
137 const unsigned int tile_width = tile_size >> 16;
138 const unsigned int tile_height = tile_size & 65535;
139 const uint32_t output_frame_width_in_tiles =
140 output_frame_width / tile_width;
141
142 // Copy the tile to the output frame.
143 const int row_offset =
144 (*tile_idx / output_frame_width_in_tiles) * tile_height;
145 const int col_offset =
146 (*tile_idx % output_frame_width_in_tiles) * tile_width;
147
148 aom_img_copy_tile(img, output, row_offset, col_offset);
149 (*tile_idx)++;
150 }
151 }
152
img_write_to_file(const aom_image_t * img,FILE * file,int output_format)153 static void img_write_to_file(const aom_image_t *img, FILE *file,
154 int output_format) {
155 if (output_format == YUV)
156 aom_img_write(img, file);
157 else if (output_format == NV12)
158 aom_img_write_nv12(img, file);
159 else
160 die("Invalid output format");
161 }
162
main(int argc,char ** argv)163 int main(int argc, char **argv) {
164 FILE *outfile = NULL;
165 AvxVideoReader *reader = NULL;
166 const AvxVideoInfo *info = NULL;
167 int num_references;
168 aom_img_fmt_t ref_fmt = 0;
169 aom_image_t reference_images[MAX_EXTERNAL_REFERENCES];
170 aom_image_t output;
171 aom_image_t *output_shifted = NULL;
172 size_t frame_size = 0;
173 const unsigned char *frame = NULL;
174 int i, j;
175 const char *tile_list_file = NULL;
176 int output_format = YUV1D;
177 exec_name = argv[0];
178
179 if (argc < 5) die("Invalid number of arguments.");
180
181 reader = aom_video_reader_open(argv[1]);
182 if (!reader) die("Failed to open %s for reading.", argv[1]);
183
184 if (!(outfile = fopen(argv[2], "wb")))
185 die("Failed to open %s for writing.", argv[2]);
186
187 num_references = (int)strtol(argv[3], NULL, 0);
188 tile_list_file = argv[4];
189
190 if (argc > 5) output_format = (int)strtol(argv[5], NULL, 0);
191 if (output_format < YUV1D || output_format > NV12)
192 die("Output format out of range [0, 2]");
193
194 info = aom_video_reader_get_info(reader);
195
196 aom_codec_iface_t *decoder;
197 if (info->codec_fourcc == LST_FOURCC)
198 decoder = get_aom_decoder_by_fourcc(AV1_FOURCC);
199 else
200 die("Unknown input codec.");
201 printf("Using %s\n", aom_codec_iface_name(decoder));
202
203 aom_codec_ctx_t codec;
204 if (aom_codec_dec_init(&codec, decoder, NULL, 0))
205 die_codec(&codec, "Failed to initialize decoder.");
206
207 if (AOM_CODEC_CONTROL_TYPECHECKED(&codec, AV1D_SET_IS_ANNEXB,
208 info->is_annexb)) {
209 die("Failed to set annex b status");
210 }
211
212 // Decode anchor frames.
213 AOM_CODEC_CONTROL_TYPECHECKED(&codec, AV1_SET_TILE_MODE, 0);
214 for (i = 0; i < num_references; ++i) {
215 aom_video_reader_read_frame(reader);
216 frame = aom_video_reader_get_frame(reader, &frame_size);
217 if (aom_codec_decode(&codec, frame, frame_size, NULL))
218 die_codec(&codec, "Failed to decode frame.");
219
220 if (i == 0) {
221 if (AOM_CODEC_CONTROL_TYPECHECKED(&codec, AV1D_GET_IMG_FORMAT, &ref_fmt))
222 die_codec(&codec, "Failed to get the image format");
223
224 int frame_res[2];
225 if (AOM_CODEC_CONTROL_TYPECHECKED(&codec, AV1D_GET_FRAME_SIZE, frame_res))
226 die_codec(&codec, "Failed to get the image frame size");
227
228 // Allocate memory to store decoded references. Allocate memory with the
229 // border so that it can be used as a reference.
230 for (j = 0; j < num_references; j++) {
231 unsigned int border = AOM_DEC_BORDER_IN_PIXELS;
232 if (!aom_img_alloc_with_border(&reference_images[j], ref_fmt,
233 frame_res[0], frame_res[1], 32, 8,
234 border)) {
235 die("Failed to allocate references.");
236 }
237 }
238 }
239
240 if (AOM_CODEC_CONTROL_TYPECHECKED(&codec, AV1_COPY_NEW_FRAME_IMAGE,
241 &reference_images[i]))
242 die_codec(&codec, "Failed to copy decoded reference frame");
243
244 aom_codec_iter_t iter = NULL;
245 aom_image_t *img = NULL;
246 while ((img = aom_codec_get_frame(&codec, &iter)) != NULL) {
247 char name[1024];
248 snprintf(name, sizeof(name), "ref_%d.yuv", i);
249 printf("writing ref image to %s, %u, %u\n", name, img->d_w, img->d_h);
250 FILE *ref_file = fopen(name, "wb");
251 aom_img_write(img, ref_file);
252 fclose(ref_file);
253 }
254 }
255
256 FILE *infile = aom_video_reader_get_file(reader);
257 // Record the offset of the first camera image.
258 const FileOffset camera_frame_pos = ftello(infile);
259
260 printf("Loading compressed frames into memory.\n");
261
262 // Count the frames in the lightfield.
263 int num_frames = 0;
264 while (aom_video_reader_read_frame(reader)) {
265 ++num_frames;
266 }
267 if (num_frames < 1) die("Input light field has no frames.");
268
269 // Read all of the lightfield frames into memory.
270 unsigned char **frames =
271 (unsigned char **)malloc(num_frames * sizeof(unsigned char *));
272 size_t *frame_sizes = (size_t *)malloc(num_frames * sizeof(size_t));
273 if (!(frames && frame_sizes)) die("Failed to allocate frame data.");
274 // Seek to the first camera image.
275 fseeko(infile, camera_frame_pos, SEEK_SET);
276 for (int f = 0; f < num_frames; ++f) {
277 aom_video_reader_read_frame(reader);
278 frame = aom_video_reader_get_frame(reader, &frame_size);
279 frames[f] = (unsigned char *)malloc(frame_size * sizeof(unsigned char));
280 if (!frames[f]) die("Failed to allocate frame data.");
281 memcpy(frames[f], frame, frame_size);
282 frame_sizes[f] = frame_size;
283 }
284 printf("Read %d frames.\n", num_frames);
285
286 if (output_format != YUV1D) {
287 // Allocate the output frame.
288 aom_img_fmt_t out_fmt = ref_fmt;
289 if (FORCE_HIGHBITDEPTH_DECODING) out_fmt |= AOM_IMG_FMT_HIGHBITDEPTH;
290 if (!aom_img_alloc(&output, out_fmt, output_frame_width,
291 output_frame_height, 32))
292 die("Failed to allocate output image.");
293 }
294
295 printf("Decoding tile list from file.\n");
296 char line[1024];
297 FILE *tile_list_fptr = fopen(tile_list_file, "r");
298 if (!tile_list_fptr) die_codec(&codec, "Failed to open tile list file.");
299 int tile_list_cnt = 0;
300 int tile_list_writes = 0;
301 int tile_idx = 0;
302 aom_image_t *out = NULL;
303 unsigned int output_bit_depth = 0;
304
305 while ((fgets(line, 1024, tile_list_fptr)) != NULL) {
306 if (line[0] == 'F') {
307 if (output_format != YUV1D) {
308 // Write out the tile list.
309 if (tile_list_cnt) {
310 out = &output;
311 if (output_bit_depth != 0) {
312 if (!aom_shift_img(output_bit_depth, &out, &output_shifted)) {
313 die("Error allocating image");
314 }
315 }
316 img_write_to_file(out, outfile, output_format);
317 tile_list_writes++;
318 }
319
320 tile_list_cnt++;
321 tile_idx = 0;
322 // Then memset the frame.
323 memset(output.img_data, 0, output.sz);
324 }
325 continue;
326 }
327
328 int image_idx, ref_idx, tc, tr;
329 sscanf(line, "%d %d %d %d", &image_idx, &ref_idx, &tc, &tr);
330 if (image_idx >= num_frames) {
331 die("Tile list image_idx out of bounds: %d >= %d.", image_idx,
332 num_frames);
333 }
334 if (ref_idx >= num_references) {
335 die("Tile list ref_idx out of bounds: %d >= %d.", ref_idx,
336 num_references);
337 }
338 frame = frames[image_idx];
339 frame_size = frame_sizes[image_idx];
340
341 aom_image_t *img = NULL;
342 decode_tile(&codec, frame, frame_size, tr, tc, ref_idx, reference_images,
343 &output, &tile_idx, &output_bit_depth, &img, output_format);
344 if (output_format == YUV1D) {
345 out = img;
346 if (output_bit_depth != 0) {
347 if (!aom_shift_img(output_bit_depth, &out, &output_shifted)) {
348 die("Error allocating image");
349 }
350 }
351 aom_img_write(out, outfile);
352 }
353 }
354
355 if (output_format != YUV1D) {
356 // Write out the last tile list.
357 if (tile_list_writes < tile_list_cnt) {
358 out = &output;
359 if (output_bit_depth != 0) {
360 if (!aom_shift_img(output_bit_depth, &out, &output_shifted)) {
361 die("Error allocating image");
362 }
363 }
364 img_write_to_file(out, outfile, output_format);
365 }
366 }
367
368 if (output_shifted) aom_img_free(output_shifted);
369 if (output_format != YUV1D) aom_img_free(&output);
370 for (i = 0; i < num_references; i++) aom_img_free(&reference_images[i]);
371 for (int f = 0; f < num_frames; ++f) {
372 free(frames[f]);
373 }
374 free(frame_sizes);
375 free(frames);
376 if (aom_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec");
377 aom_video_reader_close(reader);
378 fclose(outfile);
379
380 return EXIT_SUCCESS;
381 }
382