FFmpeg
vf_find_rect.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2014-2015 Michael Niedermayer <michaelni@gmx.at>
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License as published by
8  * the Free Software Foundation; either version 2 of the License, or
9  * (at your option) any later version.
10  *
11  * FFmpeg is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14  * GNU General Public License for more details.
15  *
16  * You should have received a copy of the GNU General Public License along
17  * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
18  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
19  */
20 
21 /**
22  * @todo switch to dualinput
23  */
24 
25 #include "libavutil/imgutils.h"
26 #include "libavutil/opt.h"
27 #include "internal.h"
28 
29 #include "lavfutils.h"
30 
31 #define MAX_MIPMAPS 5
32 
33 typedef struct FOCContext {
34  AVClass *class;
35  float threshold;
36  int mipmaps;
37  int xmin, ymin, xmax, ymax;
38  char *obj_filename;
39  int last_x, last_y;
43  int discard;
44 } FOCContext;
45 
46 #define OFFSET(x) offsetof(FOCContext, x)
47 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
48 static const AVOption find_rect_options[] = {
49  { "object", "object bitmap filename", OFFSET(obj_filename), AV_OPT_TYPE_STRING, {.str = NULL}, .flags = FLAGS },
50  { "threshold", "set threshold", OFFSET(threshold), AV_OPT_TYPE_FLOAT, {.dbl = 0.5}, 0, 1.0, FLAGS },
51  { "mipmaps", "set mipmaps", OFFSET(mipmaps), AV_OPT_TYPE_INT, {.i64 = 3}, 1, MAX_MIPMAPS, FLAGS },
52  { "xmin", "", OFFSET(xmin), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
53  { "ymin", "", OFFSET(ymin), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
54  { "xmax", "", OFFSET(xmax), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
55  { "ymax", "", OFFSET(ymax), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
56  { "discard", "", OFFSET(discard), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, FLAGS },
57  { NULL }
58 };
59 
60 AVFILTER_DEFINE_CLASS(find_rect);
61 
63 {
64  static const enum AVPixelFormat pix_fmts[] = {
68  };
69 
71 }
72 
74 {
75  int x, y;
77  uint8_t *src, *dst;
78  if (!frame)
79  return NULL;
80 
81  frame->format = in->format;
82  frame->width = (in->width + 1) / 2;
83  frame->height = (in->height+ 1) / 2;
84 
85  if (av_frame_get_buffer(frame, 0) < 0) {
87  return NULL;
88  }
89  src = in ->data[0];
90  dst = frame->data[0];
91 
92  for(y = 0; y < frame->height; y++) {
93  for(x = 0; x < frame->width; x++) {
94  dst[x] = ( src[2*x+0]
95  + src[2*x+1]
96  + src[2*x+0 + in->linesize[0]]
97  + src[2*x+1 + in->linesize[0]]
98  + 2) >> 2;
99  }
100  src += 2*in->linesize[0];
101  dst += frame->linesize[0];
102  }
103  return frame;
104 }
105 
106 static float compare(const AVFrame *haystack, const AVFrame *obj, int offx, int offy)
107 {
108  int x,y;
109  int o_sum_v = 0;
110  int h_sum_v = 0;
111  int64_t oo_sum_v = 0;
112  int64_t hh_sum_v = 0;
113  int64_t oh_sum_v = 0;
114  float c;
115  int n = obj->height * obj->width;
116  const uint8_t *odat = obj ->data[0];
117  const uint8_t *hdat = haystack->data[0] + offx + offy * haystack->linesize[0];
118  int64_t o_sigma, h_sigma;
119 
120  for(y = 0; y < obj->height; y++) {
121  for(x = 0; x < obj->width; x++) {
122  int o_v = odat[x];
123  int h_v = hdat[x];
124  o_sum_v += o_v;
125  h_sum_v += h_v;
126  oo_sum_v += o_v * o_v;
127  hh_sum_v += h_v * h_v;
128  oh_sum_v += o_v * h_v;
129  }
130  odat += obj->linesize[0];
131  hdat += haystack->linesize[0];
132  }
133  o_sigma = n*oo_sum_v - o_sum_v*(int64_t)o_sum_v;
134  h_sigma = n*hh_sum_v - h_sum_v*(int64_t)h_sum_v;
135 
136  if (o_sigma == 0 || h_sigma == 0)
137  return 1.0;
138 
139  c = (n*oh_sum_v - o_sum_v*(int64_t)h_sum_v) / (sqrt(o_sigma)*sqrt(h_sigma));
140 
141  return 1 - fabs(c);
142 }
143 
145 {
146  AVFilterContext *ctx = inlink->dst;
147  FOCContext *foc = ctx->priv;
148 
149  if (foc->xmax <= 0)
150  foc->xmax = inlink->w - foc->obj_frame->width;
151  if (foc->ymax <= 0)
152  foc->ymax = inlink->h - foc->obj_frame->height;
153 
154  return 0;
155 }
156 
157 static float search(FOCContext *foc, int pass, int maxpass, int xmin, int xmax, int ymin, int ymax, int *best_x, int *best_y, float best_score)
158 {
159  int x, y;
160 
161  if (pass + 1 <= maxpass) {
162  int sub_x, sub_y;
163  search(foc, pass+1, maxpass, xmin>>1, (xmax+1)>>1, ymin>>1, (ymax+1)>>1, &sub_x, &sub_y, 2.0);
164  xmin = FFMAX(xmin, 2*sub_x - 4);
165  xmax = FFMIN(xmax, 2*sub_x + 4);
166  ymin = FFMAX(ymin, 2*sub_y - 4);
167  ymax = FFMIN(ymax, 2*sub_y + 4);
168  }
169 
170  for (y = ymin; y <= ymax; y++) {
171  for (x = xmin; x <= xmax; x++) {
172  float score = compare(foc->haystack_frame[pass], foc->needle_frame[pass], x, y);
173  if (score < best_score) {
174  best_score = score;
175  *best_x = x;
176  *best_y = y;
177  }
178  }
179  }
180  return best_score;
181 }
182 
184 {
185  AVFilterContext *ctx = inlink->dst;
186  FOCContext *foc = ctx->priv;
187  float best_score;
188  int best_x, best_y;
189  int i;
190  char buf[32];
191 
192  foc->haystack_frame[0] = av_frame_clone(in);
193  for (i=1; i<foc->mipmaps; i++) {
194  foc->haystack_frame[i] = downscale(foc->haystack_frame[i-1]);
195  }
196 
197  best_score = search(foc, 0, 0,
198  FFMAX(foc->xmin, foc->last_x - 8),
199  FFMIN(foc->xmax, foc->last_x + 8),
200  FFMAX(foc->ymin, foc->last_y - 8),
201  FFMIN(foc->ymax, foc->last_y + 8),
202  &best_x, &best_y, 2.0);
203 
204  best_score = search(foc, 0, foc->mipmaps - 1, foc->xmin, foc->xmax, foc->ymin, foc->ymax,
205  &best_x, &best_y, best_score);
206 
207  for (i=0; i<MAX_MIPMAPS; i++) {
209  }
210 
211  if (best_score > foc->threshold) {
212  if (foc->discard) {
213  av_frame_free(&in);
214  return 0;
215  } else {
216  return ff_filter_frame(ctx->outputs[0], in);
217  }
218  }
219 
220  av_log(ctx, AV_LOG_INFO, "Found at n=%"PRId64" pts_time=%f x=%d y=%d with score=%f\n",
221  inlink->frame_count_out, TS2D(in->pts) * av_q2d(inlink->time_base),
222  best_x, best_y, best_score);
223  foc->last_x = best_x;
224  foc->last_y = best_y;
225 
226  snprintf(buf, sizeof(buf), "%f", best_score);
227 
228  av_dict_set_int(&in->metadata, "lavfi.rect.w", foc->obj_frame->width, 0);
229  av_dict_set_int(&in->metadata, "lavfi.rect.h", foc->obj_frame->height, 0);
230  av_dict_set_int(&in->metadata, "lavfi.rect.x", best_x, 0);
231  av_dict_set_int(&in->metadata, "lavfi.rect.y", best_y, 0);
232  av_dict_set(&in->metadata, "lavfi.rect.score", buf, 0);
233 
234  return ff_filter_frame(ctx->outputs[0], in);
235 }
236 
238 {
239  FOCContext *foc = ctx->priv;
240  int i;
241 
242  for (i = 0; i < MAX_MIPMAPS; i++) {
243  av_frame_free(&foc->needle_frame[i]);
245  }
246 
247  if (foc->obj_frame)
248  av_freep(&foc->obj_frame->data[0]);
249  av_frame_free(&foc->obj_frame);
250 }
251 
253 {
254  FOCContext *foc = ctx->priv;
255  int ret, i;
256 
257  if (!foc->obj_filename) {
258  av_log(ctx, AV_LOG_ERROR, "object filename not set\n");
259  return AVERROR(EINVAL);
260  }
261 
262  foc->obj_frame = av_frame_alloc();
263  if (!foc->obj_frame)
264  return AVERROR(ENOMEM);
265 
266  if ((ret = ff_load_image(foc->obj_frame->data, foc->obj_frame->linesize,
267  &foc->obj_frame->width, &foc->obj_frame->height,
268  &foc->obj_frame->format, foc->obj_filename, ctx)) < 0)
269  return ret;
270 
271  if (foc->obj_frame->format != AV_PIX_FMT_GRAY8) {
272  av_log(ctx, AV_LOG_ERROR, "object image is not a grayscale image\n");
273  return AVERROR(EINVAL);
274  }
275 
276  foc->needle_frame[0] = av_frame_clone(foc->obj_frame);
277  for (i = 1; i < foc->mipmaps; i++) {
278  foc->needle_frame[i] = downscale(foc->needle_frame[i-1]);
279  if (!foc->needle_frame[i])
280  return AVERROR(ENOMEM);
281  }
282 
283  return 0;
284 }
285 
286 static const AVFilterPad foc_inputs[] = {
287  {
288  .name = "default",
289  .type = AVMEDIA_TYPE_VIDEO,
290  .config_props = config_input,
291  .filter_frame = filter_frame,
292  },
293 };
294 
295 static const AVFilterPad foc_outputs[] = {
296  {
297  .name = "default",
298  .type = AVMEDIA_TYPE_VIDEO,
299  },
300 };
301 
303  .name = "find_rect",
304  .description = NULL_IF_CONFIG_SMALL("Find a user specified object."),
305  .priv_size = sizeof(FOCContext),
306  .init = init,
307  .uninit = uninit,
311  .priv_class = &find_rect_class,
312 };
FOCContext::threshold
float threshold
Definition: vf_find_rect.c:35
AVPixelFormat
AVPixelFormat
Pixel format.
Definition: pixfmt.h:64
AVERROR
Filter the word “frame” indicates either a video frame or a group of audio as stored in an AVFrame structure Format for each input and each output the list of supported formats For video that means pixel format For audio that means channel sample they are references to shared objects When the negotiation mechanism computes the intersection of the formats supported at each end of a all references to both lists are replaced with a reference to the intersection And when a single format is eventually chosen for a link amongst the remaining all references to the list are updated That means that if a filter requires that its input and output have the same format amongst a supported all it has to do is use a reference to the same list of formats query_formats can leave some formats unset and return AVERROR(EAGAIN) to cause the negotiation mechanism toagain later. That can be used by filters with complex requirements to use the format negotiated on one link to set the formats supported on another. Frame references ownership and permissions
opt.h
FLAGS
#define FLAGS
Definition: vf_find_rect.c:47
av_frame_get_buffer
int av_frame_get_buffer(AVFrame *frame, int align)
Allocate new buffer(s) for audio or video data.
Definition: frame.c:246
ff_filter_frame
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
Definition: avfilter.c:1017
inlink
The exact code depends on how similar the blocks are and how related they are to the and needs to apply these operations to the correct inlink or outlink if there are several Macros are available to factor that when no extra processing is inlink
Definition: filter_design.txt:212
av_frame_free
void av_frame_free(AVFrame **frame)
Free the frame and any dynamically allocated objects in it, e.g.
Definition: frame.c:112
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:303
AVFrame::pts
int64_t pts
Presentation timestamp in time_base units (time when frame should be shown to user).
Definition: frame.h:396
AVFrame::width
int width
Definition: frame.h:361
downscale
static AVFrame * downscale(AVFrame *in)
Definition: vf_find_rect.c:73
AVOption
AVOption.
Definition: opt.h:247
FOCContext::obj_filename
char * obj_filename
Definition: vf_find_rect.c:38
FFMAX
#define FFMAX(a, b)
Definition: macros.h:47
AVFilter::name
const char * name
Filter name.
Definition: avfilter.h:153
FOCContext::last_x
int last_x
Definition: vf_find_rect.c:39
AVFrame::data
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
Definition: frame.h:317
FOCContext::xmin
int xmin
Definition: vf_find_rect.c:37
AVFilterPad
A filter pad used for either input or output.
Definition: internal.h:50
av_frame_alloc
AVFrame * av_frame_alloc(void)
Allocate an AVFrame and set its fields to default values.
Definition: frame.c:99
AV_LOG_ERROR
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:180
av_cold
#define av_cold
Definition: attributes.h:90
filter_frame
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
Definition: vf_find_rect.c:183
av_q2d
static double av_q2d(AVRational a)
Convert an AVRational to a double.
Definition: rational.h:104
ff_set_common_formats_from_list
int ff_set_common_formats_from_list(AVFilterContext *ctx, const int *fmts)
Equivalent to ff_set_common_formats(ctx, ff_make_format_list(fmts))
Definition: formats.c:703
query_formats
static int query_formats(AVFilterContext *ctx)
Definition: vf_find_rect.c:62
FOCContext::discard
int discard
Definition: vf_find_rect.c:43
pix_fmts
static enum AVPixelFormat pix_fmts[]
Definition: libkvazaar.c:290
ctx
AVFormatContext * ctx
Definition: movenc.c:48
av_frame_clone
AVFrame * av_frame_clone(const AVFrame *src)
Create a new frame that references the same data as src.
Definition: frame.c:424
AV_PIX_FMT_YUV420P
@ AV_PIX_FMT_YUV420P
planar YUV 4:2:0, 12bpp, (1 Cr & Cb sample per 2x2 Y samples)
Definition: pixfmt.h:66
FOCContext::ymax
int ymax
Definition: vf_find_rect.c:37
pass
#define pass
Definition: fft_template.c:601
FILTER_INPUTS
#define FILTER_INPUTS(array)
Definition: internal.h:152
AVClass
Describe the class of an AVClass context structure.
Definition: log.h:66
fabs
static __device__ float fabs(float a)
Definition: cuda_runtime.h:182
TS2D
#define TS2D(ts)
Definition: internal.h:224
NULL
#define NULL
Definition: coverity.c:32
AV_PIX_FMT_YUVJ420P
@ AV_PIX_FMT_YUVJ420P
planar YUV 4:2:0, 12bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV420P and setting col...
Definition: pixfmt.h:78
src
#define src
Definition: vp8dsp.c:255
AVFILTER_DEFINE_CLASS
AVFILTER_DEFINE_CLASS(find_rect)
AV_PIX_FMT_GRAY8
@ AV_PIX_FMT_GRAY8
Y , 8bpp.
Definition: pixfmt.h:74
c
Undefined Behavior In the C some operations are like signed integer dereferencing freed accessing outside allocated Undefined Behavior must not occur in a C it is not safe even if the output of undefined operations is unused The unsafety may seem nit picking but Optimizing compilers have in fact optimized code on the assumption that no undefined Behavior occurs Optimizing code based on wrong assumptions can and has in some cases lead to effects beyond the output of computations The signed integer overflow problem in speed critical code Code which is highly optimized and works with signed integers sometimes has the problem that often the output of the computation does not c
Definition: undefined.txt:32
init
static av_cold int init(AVFilterContext *ctx)
Definition: vf_find_rect.c:252
FOCContext::mipmaps
int mipmaps
Definition: vf_find_rect.c:36
NULL_IF_CONFIG_SMALL
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification.
Definition: internal.h:117
foc_inputs
static const AVFilterPad foc_inputs[]
Definition: vf_find_rect.c:286
FOCContext
Definition: vf_find_rect.c:33
ff_load_image
int ff_load_image(uint8_t *data[4], int linesize[4], int *w, int *h, enum AVPixelFormat *pix_fmt, const char *filename, void *log_ctx)
Load image from filename and put the resulting image in data.
Definition: lavfutils.c:34
AVFrame::format
int format
format of the frame, -1 if unknown or unset Values correspond to enum AVPixelFormat for video frames,...
Definition: frame.h:376
FOCContext::haystack_frame
AVFrame * haystack_frame[MAX_MIPMAPS]
Definition: vf_find_rect.c:42
OFFSET
#define OFFSET(x)
Definition: vf_find_rect.c:46
ff_vf_find_rect
const AVFilter ff_vf_find_rect
Definition: vf_find_rect.c:302
AV_LOG_INFO
#define AV_LOG_INFO
Standard information.
Definition: log.h:191
internal.h
AV_OPT_TYPE_FLOAT
@ AV_OPT_TYPE_FLOAT
Definition: opt.h:227
i
int i
Definition: input.c:406
config_input
static int config_input(AVFilterLink *inlink)
Definition: vf_find_rect.c:144
FOCContext::last_y
int last_y
Definition: vf_find_rect.c:39
FFMIN
#define FFMIN(a, b)
Definition: macros.h:49
AVFilterPad::name
const char * name
Pad name.
Definition: internal.h:56
FOCContext::xmax
int xmax
Definition: vf_find_rect.c:37
AVFilter
Filter definition.
Definition: avfilter.h:149
ret
ret
Definition: filter_design.txt:187
search
static float search(FOCContext *foc, int pass, int maxpass, int xmin, int xmax, int ymin, int ymax, int *best_x, int *best_y, float best_score)
Definition: vf_find_rect.c:157
frame
these buffered frames must be flushed immediately if a new input produces new the filter must not call request_frame to get more It must just process the frame or queue it The task of requesting more frames is left to the filter s request_frame method or the application If a filter has several the filter must be ready for frames arriving randomly on any input any filter with several inputs will most likely require some kind of queuing mechanism It is perfectly acceptable to have a limited queue and to drop frames when the inputs are too unbalanced request_frame For filters that do not use the this method is called when a frame is wanted on an output For a it should directly call filter_frame on the corresponding output For a if there are queued frames already one of these frames should be pushed If the filter should request a frame on one of its repeatedly until at least one frame has been pushed Return or at least make progress towards producing a frame
Definition: filter_design.txt:264
FOCContext::needle_frame
AVFrame * needle_frame[MAX_MIPMAPS]
Definition: vf_find_rect.c:41
FOCContext::ymin
int ymin
Definition: vf_find_rect.c:37
AVFrame::height
int height
Definition: frame.h:361
compare
static float compare(const AVFrame *haystack, const AVFrame *obj, int offx, int offy)
Definition: vf_find_rect.c:106
AV_PIX_FMT_NONE
@ AV_PIX_FMT_NONE
Definition: pixfmt.h:65
AV_OPT_TYPE_INT
@ AV_OPT_TYPE_INT
Definition: opt.h:224
AVFrame::metadata
AVDictionary * metadata
metadata.
Definition: frame.h:572
uninit
static av_cold void uninit(AVFilterContext *ctx)
Definition: vf_find_rect.c:237
find_rect_options
static const AVOption find_rect_options[]
Definition: vf_find_rect.c:48
foc_outputs
static const AVFilterPad foc_outputs[]
Definition: vf_find_rect.c:295
lavfutils.h
MAX_MIPMAPS
#define MAX_MIPMAPS
Definition: vf_find_rect.c:31
AVFilterContext
An instance of a filter.
Definition: avfilter.h:346
av_dict_set_int
int av_dict_set_int(AVDictionary **pm, const char *key, int64_t value, int flags)
Convenience wrapper for av_dict_set that converts the value to a string and stores it.
Definition: dict.c:147
AVMEDIA_TYPE_VIDEO
@ AVMEDIA_TYPE_VIDEO
Definition: avutil.h:201
AV_OPT_TYPE_BOOL
@ AV_OPT_TYPE_BOOL
Definition: opt.h:241
FILTER_OUTPUTS
#define FILTER_OUTPUTS(array)
Definition: internal.h:153
av_freep
#define av_freep(p)
Definition: tableprint_vlc.h:35
av_dict_set
int av_dict_set(AVDictionary **pm, const char *key, const char *value, int flags)
Set the given entry in *pm, overwriting an existing entry.
Definition: dict.c:70
FOCContext::obj_frame
AVFrame * obj_frame
Definition: vf_find_rect.c:40
imgutils.h
AVFrame::linesize
int linesize[AV_NUM_DATA_POINTERS]
For video, size in bytes of each picture line.
Definition: frame.h:334
av_log
#define av_log(a,...)
Definition: tableprint_vlc.h:28
AV_OPT_TYPE_STRING
@ AV_OPT_TYPE_STRING
Definition: opt.h:228
snprintf
#define snprintf
Definition: snprintf.h:34