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 #include "video.h"
29 
30 #include "lavfutils.h"
31 
32 #define MAX_MIPMAPS 5
33 
34 typedef struct FOCContext {
35  AVClass *class;
36  float threshold;
37  int mipmaps;
38  int xmin, ymin, xmax, ymax;
39  char *obj_filename;
40  int last_x, last_y;
44  int discard;
45 } FOCContext;
46 
47 #define OFFSET(x) offsetof(FOCContext, x)
48 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
49 static const AVOption find_rect_options[] = {
50  { "object", "object bitmap filename", OFFSET(obj_filename), AV_OPT_TYPE_STRING, {.str = NULL}, .flags = FLAGS },
51  { "threshold", "set threshold", OFFSET(threshold), AV_OPT_TYPE_FLOAT, {.dbl = 0.5}, 0, 1.0, FLAGS },
52  { "mipmaps", "set mipmaps", OFFSET(mipmaps), AV_OPT_TYPE_INT, {.i64 = 3}, 1, MAX_MIPMAPS, FLAGS },
53  { "xmin", "", OFFSET(xmin), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
54  { "ymin", "", OFFSET(ymin), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
55  { "xmax", "", OFFSET(xmax), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
56  { "ymax", "", OFFSET(ymax), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
57  { "discard", "", OFFSET(discard), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, FLAGS },
58  { NULL }
59 };
60 
61 AVFILTER_DEFINE_CLASS(find_rect);
62 
64 {
65  int x, y;
67  uint8_t *src, *dst;
68  if (!frame)
69  return NULL;
70 
71  frame->format = in->format;
72  frame->width = (in->width + 1) / 2;
73  frame->height = (in->height+ 1) / 2;
74 
75  if (av_frame_get_buffer(frame, 0) < 0) {
77  return NULL;
78  }
79  src = in ->data[0];
80  dst = frame->data[0];
81 
82  for(y = 0; y < frame->height; y++) {
83  for(x = 0; x < frame->width; x++) {
84  dst[x] = ( src[2*x+0]
85  + src[2*x+1]
86  + src[2*x+0 + in->linesize[0]]
87  + src[2*x+1 + in->linesize[0]]
88  + 2) >> 2;
89  }
90  src += 2*in->linesize[0];
91  dst += frame->linesize[0];
92  }
93  return frame;
94 }
95 
96 static float compare(const AVFrame *haystack, const AVFrame *obj, int offx, int offy)
97 {
98  int x,y;
99  int o_sum_v = 0;
100  int h_sum_v = 0;
101  int64_t oo_sum_v = 0;
102  int64_t hh_sum_v = 0;
103  int64_t oh_sum_v = 0;
104  float c;
105  int n = obj->height * obj->width;
106  const uint8_t *odat = obj ->data[0];
107  const uint8_t *hdat = haystack->data[0] + offx + offy * haystack->linesize[0];
108  int64_t o_sigma, h_sigma;
109 
110  for(y = 0; y < obj->height; y++) {
111  for(x = 0; x < obj->width; x++) {
112  int o_v = odat[x];
113  int h_v = hdat[x];
114  o_sum_v += o_v;
115  h_sum_v += h_v;
116  oo_sum_v += o_v * o_v;
117  hh_sum_v += h_v * h_v;
118  oh_sum_v += o_v * h_v;
119  }
120  odat += obj->linesize[0];
121  hdat += haystack->linesize[0];
122  }
123  o_sigma = n*oo_sum_v - o_sum_v*(int64_t)o_sum_v;
124  h_sigma = n*hh_sum_v - h_sum_v*(int64_t)h_sum_v;
125 
126  if (o_sigma == 0 || h_sigma == 0)
127  return 1.0;
128 
129  c = (n*oh_sum_v - o_sum_v*(int64_t)h_sum_v) / (sqrt(o_sigma)*sqrt(h_sigma));
130 
131  return 1 - fabs(c);
132 }
133 
135 {
136  AVFilterContext *ctx = inlink->dst;
137  FOCContext *foc = ctx->priv;
138 
139  if (foc->xmax <= 0)
140  foc->xmax = inlink->w - foc->obj_frame->width;
141  if (foc->ymax <= 0)
142  foc->ymax = inlink->h - foc->obj_frame->height;
143 
144  return 0;
145 }
146 
147 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)
148 {
149  int x, y;
150 
151  if (pass + 1 <= maxpass) {
152  int sub_x, sub_y;
153  search(foc, pass+1, maxpass, xmin>>1, (xmax+1)>>1, ymin>>1, (ymax+1)>>1, &sub_x, &sub_y, 2.0);
154  xmin = FFMAX(xmin, 2*sub_x - 4);
155  xmax = FFMIN(xmax, 2*sub_x + 4);
156  ymin = FFMAX(ymin, 2*sub_y - 4);
157  ymax = FFMIN(ymax, 2*sub_y + 4);
158  }
159 
160  for (y = ymin; y <= ymax; y++) {
161  for (x = xmin; x <= xmax; x++) {
162  float score = compare(foc->haystack_frame[pass], foc->needle_frame[pass], x, y);
163  if (score < best_score) {
164  best_score = score;
165  *best_x = x;
166  *best_y = y;
167  }
168  }
169  }
170  return best_score;
171 }
172 
174 {
175  AVFilterContext *ctx = inlink->dst;
176  FOCContext *foc = ctx->priv;
177  float best_score;
178  int best_x, best_y;
179  int i;
180  char buf[32];
181 
182  foc->haystack_frame[0] = av_frame_clone(in);
183  for (i=1; i<foc->mipmaps; i++) {
184  foc->haystack_frame[i] = downscale(foc->haystack_frame[i-1]);
185  }
186 
187  best_score = search(foc, 0, 0,
188  FFMAX(foc->xmin, foc->last_x - 8),
189  FFMIN(foc->xmax, foc->last_x + 8),
190  FFMAX(foc->ymin, foc->last_y - 8),
191  FFMIN(foc->ymax, foc->last_y + 8),
192  &best_x, &best_y, 2.0);
193 
194  best_score = search(foc, 0, foc->mipmaps - 1, foc->xmin, foc->xmax, foc->ymin, foc->ymax,
195  &best_x, &best_y, best_score);
196 
197  for (i=0; i<MAX_MIPMAPS; i++) {
199  }
200 
201  if (best_score > foc->threshold) {
202  if (foc->discard) {
203  av_frame_free(&in);
204  return 0;
205  } else {
206  return ff_filter_frame(ctx->outputs[0], in);
207  }
208  }
209 
210  av_log(ctx, AV_LOG_INFO, "Found at n=%"PRId64" pts_time=%f x=%d y=%d with score=%f\n",
211  inlink->frame_count_out, TS2D(in->pts) * av_q2d(inlink->time_base),
212  best_x, best_y, best_score);
213  foc->last_x = best_x;
214  foc->last_y = best_y;
215 
216  snprintf(buf, sizeof(buf), "%f", best_score);
217 
218  av_dict_set_int(&in->metadata, "lavfi.rect.w", foc->obj_frame->width, 0);
219  av_dict_set_int(&in->metadata, "lavfi.rect.h", foc->obj_frame->height, 0);
220  av_dict_set_int(&in->metadata, "lavfi.rect.x", best_x, 0);
221  av_dict_set_int(&in->metadata, "lavfi.rect.y", best_y, 0);
222  av_dict_set(&in->metadata, "lavfi.rect.score", buf, 0);
223 
224  return ff_filter_frame(ctx->outputs[0], in);
225 }
226 
228 {
229  FOCContext *foc = ctx->priv;
230  int i;
231 
232  for (i = 0; i < MAX_MIPMAPS; i++) {
233  av_frame_free(&foc->needle_frame[i]);
235  }
236 
237  if (foc->obj_frame)
238  av_freep(&foc->obj_frame->data[0]);
239  av_frame_free(&foc->obj_frame);
240 }
241 
243 {
244  FOCContext *foc = ctx->priv;
245  int ret, i;
246 
247  if (!foc->obj_filename) {
248  av_log(ctx, AV_LOG_ERROR, "object filename not set\n");
249  return AVERROR(EINVAL);
250  }
251 
252  foc->obj_frame = av_frame_alloc();
253  if (!foc->obj_frame)
254  return AVERROR(ENOMEM);
255 
256  if ((ret = ff_load_image(foc->obj_frame->data, foc->obj_frame->linesize,
257  &foc->obj_frame->width, &foc->obj_frame->height,
258  &foc->obj_frame->format, foc->obj_filename, ctx)) < 0)
259  return ret;
260 
261  if (foc->obj_frame->format != AV_PIX_FMT_GRAY8) {
262  av_log(ctx, AV_LOG_ERROR, "object image is not a grayscale image\n");
263  return AVERROR(EINVAL);
264  }
265 
266  foc->needle_frame[0] = av_frame_clone(foc->obj_frame);
267  for (i = 1; i < foc->mipmaps; i++) {
268  foc->needle_frame[i] = downscale(foc->needle_frame[i-1]);
269  if (!foc->needle_frame[i])
270  return AVERROR(ENOMEM);
271  }
272 
273  return 0;
274 }
275 
276 static const AVFilterPad foc_inputs[] = {
277  {
278  .name = "default",
279  .type = AVMEDIA_TYPE_VIDEO,
280  .config_props = config_input,
281  .filter_frame = filter_frame,
282  },
283 };
284 
286  .name = "find_rect",
287  .description = NULL_IF_CONFIG_SMALL("Find a user specified object."),
288  .priv_size = sizeof(FOCContext),
289  .init = init,
290  .uninit = uninit,
295  .priv_class = &find_rect_class,
296 };
FOCContext::threshold
float threshold
Definition: vf_find_rect.c:36
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:48
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:288
ff_filter_frame
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
Definition: avfilter.c:1018
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:160
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:375
AVFrame::pts
int64_t pts
Presentation timestamp in time_base units (time when frame should be shown to user).
Definition: frame.h:487
AVFrame::width
int width
Definition: frame.h:447
downscale
static AVFrame * downscale(AVFrame *in)
Definition: vf_find_rect.c:63
AVOption
AVOption.
Definition: opt.h:346
FOCContext::obj_filename
char * obj_filename
Definition: vf_find_rect.c:39
FFMAX
#define FFMAX(a, b)
Definition: macros.h:47
AVFilter::name
const char * name
Filter name.
Definition: avfilter.h:170
video.h
FOCContext::last_x
int last_x
Definition: vf_find_rect.c:40
AVFrame::data
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
Definition: frame.h:396
FOCContext::xmin
int xmin
Definition: vf_find_rect.c:38
AVFilterPad
A filter pad used for either input or output.
Definition: internal.h:33
av_frame_alloc
AVFrame * av_frame_alloc(void)
Allocate an AVFrame and set its fields to default values.
Definition: frame.c:148
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
ff_video_default_filterpad
const AVFilterPad ff_video_default_filterpad[1]
An AVFilterPad array whose only entry has name "default" and is of type AVMEDIA_TYPE_VIDEO.
Definition: video.c:37
filter_frame
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
Definition: vf_find_rect.c:173
av_q2d
static double av_q2d(AVRational a)
Convert an AVRational to a double.
Definition: rational.h:104
FOCContext::discard
int discard
Definition: vf_find_rect.c:44
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:593
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:73
FOCContext::ymax
int ymax
Definition: vf_find_rect.c:38
FILTER_INPUTS
#define FILTER_INPUTS(array)
Definition: internal.h:182
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:258
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:85
AVFILTER_DEFINE_CLASS
AVFILTER_DEFINE_CLASS(find_rect)
AV_PIX_FMT_GRAY8
@ AV_PIX_FMT_GRAY8
Y , 8bpp.
Definition: pixfmt.h:81
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:242
FOCContext::mipmaps
int mipmaps
Definition: vf_find_rect.c:37
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:106
foc_inputs
static const AVFilterPad foc_inputs[]
Definition: vf_find_rect.c:276
FOCContext
Definition: vf_find_rect.c:34
FILTER_PIXFMTS
#define FILTER_PIXFMTS(...)
Definition: internal.h:168
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:462
FOCContext::haystack_frame
AVFrame * haystack_frame[MAX_MIPMAPS]
Definition: vf_find_rect.c:43
OFFSET
#define OFFSET(x)
Definition: vf_find_rect.c:47
ff_vf_find_rect
const AVFilter ff_vf_find_rect
Definition: vf_find_rect.c:285
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:238
i
#define i(width, name, range_min, range_max)
Definition: cbs_h2645.c:255
config_input
static int config_input(AVFilterLink *inlink)
Definition: vf_find_rect.c:134
FOCContext::last_y
int last_y
Definition: vf_find_rect.c:40
FFMIN
#define FFMIN(a, b)
Definition: macros.h:49
AVFilterPad::name
const char * name
Pad name.
Definition: internal.h:39
FOCContext::xmax
int xmax
Definition: vf_find_rect.c:38
AVFilter
Filter definition.
Definition: avfilter.h:166
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:147
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:42
FOCContext::ymin
int ymin
Definition: vf_find_rect.c:38
AVFrame::height
int height
Definition: frame.h:447
compare
static float compare(const AVFrame *haystack, const AVFrame *obj, int offx, int offy)
Definition: vf_find_rect.c:96
AV_OPT_TYPE_INT
@ AV_OPT_TYPE_INT
Definition: opt.h:235
AVFrame::metadata
AVDictionary * metadata
metadata.
Definition: frame.h:693
uninit
static av_cold void uninit(AVFilterContext *ctx)
Definition: vf_find_rect.c:227
AVFILTER_FLAG_METADATA_ONLY
#define AVFILTER_FLAG_METADATA_ONLY
The filter is a "metadata" filter - it does not modify the frame data in any way.
Definition: avfilter.h:133
find_rect_options
static const AVOption find_rect_options[]
Definition: vf_find_rect.c:49
lavfutils.h
MAX_MIPMAPS
#define MAX_MIPMAPS
Definition: vf_find_rect.c:32
AVFilterContext
An instance of a filter.
Definition: avfilter.h:407
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:167
AVMEDIA_TYPE_VIDEO
@ AVMEDIA_TYPE_VIDEO
Definition: avutil.h:201
AV_OPT_TYPE_BOOL
@ AV_OPT_TYPE_BOOL
Definition: opt.h:251
FILTER_OUTPUTS
#define FILTER_OUTPUTS(array)
Definition: internal.h:183
av_freep
#define av_freep(p)
Definition: tableprint_vlc.h:34
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:88
src
INIT_CLIP pixel * src
Definition: h264pred_template.c:418
FOCContext::obj_frame
AVFrame * obj_frame
Definition: vf_find_rect.c:41
imgutils.h
flags
#define flags(name, subs,...)
Definition: cbs_av1.c:474
AVFrame::linesize
int linesize[AV_NUM_DATA_POINTERS]
For video, a positive or negative value, which is typically indicating the size in bytes of each pict...
Definition: frame.h:420
av_log
#define av_log(a,...)
Definition: tableprint_vlc.h:27
AV_OPT_TYPE_STRING
@ AV_OPT_TYPE_STRING
Definition: opt.h:239
snprintf
#define snprintf
Definition: snprintf.h:34