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/opt.h"
26 
27 #include "filters.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 = INT_MAX}, 0, INT_MAX, FLAGS },
56  { "ymax", "", OFFSET(ymax), AV_OPT_TYPE_INT, {.i64 = INT_MAX}, 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  int w2 = in->width/2;
83  int h2 = in->height/2;
84  for(y = 0; y < h2; y++) {
85  for(x = 0; x < w2; x++) {
86  dst[x] = ( src[2*x+0]
87  + src[2*x+1]
88  + src[2*x+0 + in->linesize[0]]
89  + src[2*x+1 + in->linesize[0]]
90  + 2) >> 2;
91  }
92  src += 2*in->linesize[0];
93  dst += frame->linesize[0];
94  }
95  src = in ->data[0];
96  dst = frame->data[0];
97  for(y = 0; y < frame->height; y++) {
98  int yd = y < h2 ? in->linesize[0] : 0;
99  x = yd ? w2 : 0;
100  for(; x < frame->width; x++) {
101  dst[x] = ( src[2*x+0]
102  + src[FFMIN(2*x+1, w2)]
103  + src[2*x+0 + yd]
104  + src[FFMIN(2*x+1, w2) + yd]
105  + 2) >> 2;
106  }
107  src += 2*in->linesize[0];
108  dst += frame->linesize[0];
109  }
110 
111  return frame;
112 }
113 
114 static float compare(const AVFrame *haystack, const AVFrame *obj, int offx, int offy)
115 {
116  int x,y;
117  int o_sum_v = 0;
118  int h_sum_v = 0;
119  int64_t oo_sum_v = 0;
120  int64_t hh_sum_v = 0;
121  int64_t oh_sum_v = 0;
122  float c;
123  int n = obj->height * obj->width;
124  const uint8_t *odat = obj ->data[0];
125  const uint8_t *hdat = haystack->data[0] + offx + offy * haystack->linesize[0];
126  int64_t o_sigma, h_sigma;
127 
128  for(y = 0; y < obj->height; y++) {
129  for(x = 0; x < obj->width; x++) {
130  int o_v = odat[x];
131  int h_v = hdat[x];
132  o_sum_v += o_v;
133  h_sum_v += h_v;
134  oo_sum_v += o_v * o_v;
135  hh_sum_v += h_v * h_v;
136  oh_sum_v += o_v * h_v;
137  }
138  odat += obj->linesize[0];
139  hdat += haystack->linesize[0];
140  }
141  o_sigma = n*oo_sum_v - o_sum_v*(int64_t)o_sum_v;
142  h_sigma = n*hh_sum_v - h_sum_v*(int64_t)h_sum_v;
143 
144  if (o_sigma == 0 || h_sigma == 0)
145  return 1.0;
146 
147  c = (n*oh_sum_v - o_sum_v*(int64_t)h_sum_v) / (sqrt(o_sigma)*sqrt(h_sigma));
148 
149  return 1 - fabs(c);
150 }
151 
152 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)
153 {
154  int x, y;
155 
156  if (pass + 1 <= maxpass) {
157  int sub_x, sub_y;
158  search(foc, pass+1, maxpass, xmin>>1, (xmax+1)>>1, ymin>>1, (ymax+1)>>1, &sub_x, &sub_y, 2.0);
159  xmin = FFMAX(xmin, 2*sub_x - 4);
160  xmax = FFMIN(xmax, 2*sub_x + 4);
161  ymin = FFMAX(ymin, 2*sub_y - 4);
162  ymax = FFMIN(ymax, 2*sub_y + 4);
163  }
164 
165  for (y = ymin; y <= ymax; y++) {
166  for (x = xmin; x <= xmax; x++) {
167  float score = compare(foc->haystack_frame[pass], foc->needle_frame[pass], x, y);
168  if (score < best_score) {
169  best_score = score;
170  *best_x = x;
171  *best_y = y;
172  }
173  }
174  }
175  return best_score;
176 }
177 
179 {
181  AVFilterContext *ctx = inlink->dst;
182  FOCContext *foc = ctx->priv;
183  float best_score;
184  int best_x, best_y;
185  int i;
186  char buf[32];
187 
188  int xmin = FFMAX(foc->xmin, 0);
189  int ymin = FFMAX(foc->ymin, 0);
190  int xmax = FFMIN(foc->xmax, inlink->w - foc->obj_frame->width );
191  int ymax = FFMIN(foc->ymax, inlink->h - foc->obj_frame->height);
192 
193  if (xmin > xmax || ymin > ymax)
194  av_log(ctx, AV_LOG_WARNING, "x/y min/max are invalid for the current frame\n");
195 
196 
197  foc->haystack_frame[0] = av_frame_clone(in);
198  for (i=1; i<foc->mipmaps; i++) {
199  foc->haystack_frame[i] = downscale(foc->haystack_frame[i-1]);
200  }
201 
202  best_score = search(foc, 0, 0,
203  FFMAX(xmin, foc->last_x - 8),
204  FFMIN(xmax, foc->last_x + 8),
205  FFMAX(ymin, foc->last_y - 8),
206  FFMIN(ymax, foc->last_y + 8),
207  &best_x, &best_y, 2.0);
208 
209  best_score = search(foc, 0, foc->mipmaps - 1, xmin, xmax, ymin, ymax,
210  &best_x, &best_y, best_score);
211 
212  for (i=0; i<MAX_MIPMAPS; i++) {
214  }
215 
216  if (best_score > foc->threshold) {
217  if (foc->discard) {
218  av_frame_free(&in);
219  return 0;
220  } else {
221  return ff_filter_frame(ctx->outputs[0], in);
222  }
223  }
224 
225  av_log(ctx, AV_LOG_INFO, "Found at n=%"PRId64" pts_time=%f x=%d y=%d with score=%f\n",
226  inl->frame_count_out, TS2D(in->pts) * av_q2d(inlink->time_base),
227  best_x, best_y, best_score);
228  foc->last_x = best_x;
229  foc->last_y = best_y;
230 
231  snprintf(buf, sizeof(buf), "%f", best_score);
232 
233  av_dict_set_int(&in->metadata, "lavfi.rect.w", foc->obj_frame->width, 0);
234  av_dict_set_int(&in->metadata, "lavfi.rect.h", foc->obj_frame->height, 0);
235  av_dict_set_int(&in->metadata, "lavfi.rect.x", best_x, 0);
236  av_dict_set_int(&in->metadata, "lavfi.rect.y", best_y, 0);
237  av_dict_set(&in->metadata, "lavfi.rect.score", buf, 0);
238 
239  return ff_filter_frame(ctx->outputs[0], in);
240 }
241 
243 {
244  FOCContext *foc = ctx->priv;
245  int i;
246 
247  for (i = 0; i < MAX_MIPMAPS; i++) {
248  av_frame_free(&foc->needle_frame[i]);
250  }
251 
252  av_frame_free(&foc->obj_frame);
253 }
254 
256 {
257  FOCContext *foc = ctx->priv;
258  int ret, i;
259 
260  if (!foc->obj_filename) {
261  av_log(ctx, AV_LOG_ERROR, "object filename not set\n");
262  return AVERROR(EINVAL);
263  }
264 
265  ret = ff_load_image(&foc->obj_frame, foc->obj_filename, ctx);
266  if (ret < 0)
267  return ret;
268 
269  if (foc->obj_frame->format != AV_PIX_FMT_GRAY8) {
270  av_log(ctx, AV_LOG_ERROR, "object image is not a grayscale image\n");
271  return AVERROR(EINVAL);
272  }
273 
274  foc->needle_frame[0] = av_frame_clone(foc->obj_frame);
275  for (i = 1; i < foc->mipmaps; i++) {
276  foc->needle_frame[i] = downscale(foc->needle_frame[i-1]);
277  if (!foc->needle_frame[i])
278  return AVERROR(ENOMEM);
279  }
280 
281  return 0;
282 }
283 
284 static const AVFilterPad foc_inputs[] = {
285  {
286  .name = "default",
287  .type = AVMEDIA_TYPE_VIDEO,
288  .filter_frame = filter_frame,
289  },
290 };
291 
293  .p.name = "find_rect",
294  .p.description = NULL_IF_CONFIG_SMALL("Find a user specified object."),
295  .p.flags = AVFILTER_FLAG_METADATA_ONLY,
296  .p.priv_class = &find_rect_class,
297  .priv_size = sizeof(FOCContext),
298  .init = init,
299  .uninit = uninit,
303 };
AV_LOG_WARNING
#define AV_LOG_WARNING
Something somehow does not look correct.
Definition: log.h:216
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:206
ff_filter_frame
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
Definition: avfilter.c:1067
int64_t
long long int64_t
Definition: coverity.c:34
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:64
FILTER_INPUTS
#define FILTER_INPUTS(array)
Definition: filters.h:264
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:427
AVFrame::pts
int64_t pts
Presentation timestamp in time_base units (time when frame should be shown to user).
Definition: frame.h:529
AVFrame::width
int width
Definition: frame.h:499
downscale
static AVFrame * downscale(AVFrame *in)
Definition: vf_find_rect.c:63
AVOption
AVOption.
Definition: opt.h:429
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:220
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:448
FOCContext::xmin
int xmin
Definition: vf_find_rect.c:38
AVFilterPad
A filter pad used for either input or output.
Definition: filters.h:40
av_frame_alloc
AVFrame * av_frame_alloc(void)
Allocate an AVFrame and set its fields to default values.
Definition: frame.c:52
AV_LOG_ERROR
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:210
av_cold
#define av_cold
Definition: attributes.h:106
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
FFFilter
Definition: filters.h:267
filter_frame
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
Definition: vf_find_rect.c:178
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
filters.h
ctx
static AVFormatContext * ctx
Definition: movenc.c:49
av_frame_clone
AVFrame * av_frame_clone(const AVFrame *src)
Create a new frame that references the same data as src.
Definition: frame.c:483
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_OUTPUTS
#define FILTER_OUTPUTS(array)
Definition: filters.h:265
ff_load_image
int ff_load_image(struct AVFrame **outframe, const char *filename, void *log_ctx)
Load image from filename and put the resulting image in an AVFrame.
Definition: lavfutils.c:32
AVClass
Describe the class of an AVClass context structure.
Definition: log.h:76
fabs
static __device__ float fabs(float a)
Definition: cuda_runtime.h:182
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
ff_filter_link
static FilterLink * ff_filter_link(AVFilterLink *link)
Definition: filters.h:199
TS2D
#define TS2D(ts)
Definition: filters.h:482
init
static av_cold int init(AVFilterContext *ctx)
Definition: vf_find_rect.c:255
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:94
foc_inputs
static const AVFilterPad foc_inputs[]
Definition: vf_find_rect.c:284
FOCContext
Definition: vf_find_rect.c:34
ff_vf_find_rect
const FFFilter ff_vf_find_rect
Definition: vf_find_rect.c:292
dst
uint8_t ptrdiff_t const uint8_t ptrdiff_t int intptr_t intptr_t int int16_t * dst
Definition: dsp.h:87
AVFrame::format
int format
format of the frame, -1 if unknown or unset Values correspond to enum AVPixelFormat for video frames,...
Definition: frame.h:514
FOCContext::haystack_frame
AVFrame * haystack_frame[MAX_MIPMAPS]
Definition: vf_find_rect.c:43
OFFSET
#define OFFSET(x)
Definition: vf_find_rect.c:47
AV_LOG_INFO
#define AV_LOG_INFO
Standard information.
Definition: log.h:221
AV_OPT_TYPE_FLOAT
@ AV_OPT_TYPE_FLOAT
Underlying C type is float.
Definition: opt.h:271
i
#define i(width, name, range_min, range_max)
Definition: cbs_h2645.c:256
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: filters.h:46
FILTER_PIXFMTS
#define FILTER_PIXFMTS(...)
Definition: filters.h:250
FOCContext::xmax
int xmax
Definition: vf_find_rect.c:38
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:152
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:265
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:499
compare
static float compare(const AVFrame *haystack, const AVFrame *obj, int offx, int offy)
Definition: vf_find_rect.c:114
AV_OPT_TYPE_INT
@ AV_OPT_TYPE_INT
Underlying C type is int.
Definition: opt.h:259
AVFrame::metadata
AVDictionary * metadata
metadata.
Definition: frame.h:705
uninit
static av_cold void uninit(AVFilterContext *ctx)
Definition: vf_find_rect.c:242
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:183
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:274
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:177
AVMEDIA_TYPE_VIDEO
@ AVMEDIA_TYPE_VIDEO
Definition: avutil.h:200
FFFilter::p
AVFilter p
The public AVFilter.
Definition: filters.h:271
AV_OPT_TYPE_BOOL
@ AV_OPT_TYPE_BOOL
Underlying C type is int.
Definition: opt.h:327
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:86
FOCContext::obj_frame
AVFrame * obj_frame
Definition: vf_find_rect.c:41
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:472
av_log
#define av_log(a,...)
Definition: tableprint_vlc.h:27
AV_OPT_TYPE_STRING
@ AV_OPT_TYPE_STRING
Underlying C type is a uint8_t* that is either NULL or points to a C string allocated with the av_mal...
Definition: opt.h:276
snprintf
#define snprintf
Definition: snprintf.h:34
src
#define src
Definition: vp8dsp.c:248