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/mem.h"
26 #include "libavutil/opt.h"
27 
28 #include "filters.h"
29 #include "video.h"
30 
31 #include "lavfutils.h"
32 
33 #define MAX_MIPMAPS 5
34 
35 typedef struct FOCContext {
36  AVClass *class;
37  float threshold;
38  int mipmaps;
39  int xmin, ymin, xmax, ymax;
40  char *obj_filename;
41  int last_x, last_y;
45  int discard;
46 } FOCContext;
47 
48 #define OFFSET(x) offsetof(FOCContext, x)
49 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
50 static const AVOption find_rect_options[] = {
51  { "object", "object bitmap filename", OFFSET(obj_filename), AV_OPT_TYPE_STRING, {.str = NULL}, .flags = FLAGS },
52  { "threshold", "set threshold", OFFSET(threshold), AV_OPT_TYPE_FLOAT, {.dbl = 0.5}, 0, 1.0, FLAGS },
53  { "mipmaps", "set mipmaps", OFFSET(mipmaps), AV_OPT_TYPE_INT, {.i64 = 3}, 1, MAX_MIPMAPS, FLAGS },
54  { "xmin", "", OFFSET(xmin), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
55  { "ymin", "", OFFSET(ymin), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
56  { "xmax", "", OFFSET(xmax), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
57  { "ymax", "", OFFSET(ymax), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
58  { "discard", "", OFFSET(discard), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, FLAGS },
59  { NULL }
60 };
61 
62 AVFILTER_DEFINE_CLASS(find_rect);
63 
65 {
66  int x, y;
68  uint8_t *src, *dst;
69  if (!frame)
70  return NULL;
71 
72  frame->format = in->format;
73  frame->width = (in->width + 1) / 2;
74  frame->height = (in->height+ 1) / 2;
75 
76  if (av_frame_get_buffer(frame, 0) < 0) {
78  return NULL;
79  }
80  src = in ->data[0];
81  dst = frame->data[0];
82 
83  for(y = 0; y < frame->height; y++) {
84  for(x = 0; x < frame->width; x++) {
85  dst[x] = ( src[2*x+0]
86  + src[2*x+1]
87  + src[2*x+0 + in->linesize[0]]
88  + src[2*x+1 + in->linesize[0]]
89  + 2) >> 2;
90  }
91  src += 2*in->linesize[0];
92  dst += frame->linesize[0];
93  }
94  return frame;
95 }
96 
97 static float compare(const AVFrame *haystack, const AVFrame *obj, int offx, int offy)
98 {
99  int x,y;
100  int o_sum_v = 0;
101  int h_sum_v = 0;
102  int64_t oo_sum_v = 0;
103  int64_t hh_sum_v = 0;
104  int64_t oh_sum_v = 0;
105  float c;
106  int n = obj->height * obj->width;
107  const uint8_t *odat = obj ->data[0];
108  const uint8_t *hdat = haystack->data[0] + offx + offy * haystack->linesize[0];
109  int64_t o_sigma, h_sigma;
110 
111  for(y = 0; y < obj->height; y++) {
112  for(x = 0; x < obj->width; x++) {
113  int o_v = odat[x];
114  int h_v = hdat[x];
115  o_sum_v += o_v;
116  h_sum_v += h_v;
117  oo_sum_v += o_v * o_v;
118  hh_sum_v += h_v * h_v;
119  oh_sum_v += o_v * h_v;
120  }
121  odat += obj->linesize[0];
122  hdat += haystack->linesize[0];
123  }
124  o_sigma = n*oo_sum_v - o_sum_v*(int64_t)o_sum_v;
125  h_sigma = n*hh_sum_v - h_sum_v*(int64_t)h_sum_v;
126 
127  if (o_sigma == 0 || h_sigma == 0)
128  return 1.0;
129 
130  c = (n*oh_sum_v - o_sum_v*(int64_t)h_sum_v) / (sqrt(o_sigma)*sqrt(h_sigma));
131 
132  return 1 - fabs(c);
133 }
134 
136 {
137  AVFilterContext *ctx = inlink->dst;
138  FOCContext *foc = ctx->priv;
139 
140  if (foc->xmax <= 0)
141  foc->xmax = inlink->w - foc->obj_frame->width;
142  if (foc->ymax <= 0)
143  foc->ymax = inlink->h - foc->obj_frame->height;
144 
145  return 0;
146 }
147 
148 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)
149 {
150  int x, y;
151 
152  if (pass + 1 <= maxpass) {
153  int sub_x, sub_y;
154  search(foc, pass+1, maxpass, xmin>>1, (xmax+1)>>1, ymin>>1, (ymax+1)>>1, &sub_x, &sub_y, 2.0);
155  xmin = FFMAX(xmin, 2*sub_x - 4);
156  xmax = FFMIN(xmax, 2*sub_x + 4);
157  ymin = FFMAX(ymin, 2*sub_y - 4);
158  ymax = FFMIN(ymax, 2*sub_y + 4);
159  }
160 
161  for (y = ymin; y <= ymax; y++) {
162  for (x = xmin; x <= xmax; x++) {
163  float score = compare(foc->haystack_frame[pass], foc->needle_frame[pass], x, y);
164  if (score < best_score) {
165  best_score = score;
166  *best_x = x;
167  *best_y = y;
168  }
169  }
170  }
171  return best_score;
172 }
173 
175 {
177  AVFilterContext *ctx = inlink->dst;
178  FOCContext *foc = ctx->priv;
179  float best_score;
180  int best_x, best_y;
181  int i;
182  char buf[32];
183 
184  foc->haystack_frame[0] = av_frame_clone(in);
185  for (i=1; i<foc->mipmaps; i++) {
186  foc->haystack_frame[i] = downscale(foc->haystack_frame[i-1]);
187  }
188 
189  best_score = search(foc, 0, 0,
190  FFMAX(foc->xmin, foc->last_x - 8),
191  FFMIN(foc->xmax, foc->last_x + 8),
192  FFMAX(foc->ymin, foc->last_y - 8),
193  FFMIN(foc->ymax, foc->last_y + 8),
194  &best_x, &best_y, 2.0);
195 
196  best_score = search(foc, 0, foc->mipmaps - 1, foc->xmin, foc->xmax, foc->ymin, foc->ymax,
197  &best_x, &best_y, best_score);
198 
199  for (i=0; i<MAX_MIPMAPS; i++) {
201  }
202 
203  if (best_score > foc->threshold) {
204  if (foc->discard) {
205  av_frame_free(&in);
206  return 0;
207  } else {
208  return ff_filter_frame(ctx->outputs[0], in);
209  }
210  }
211 
212  av_log(ctx, AV_LOG_INFO, "Found at n=%"PRId64" pts_time=%f x=%d y=%d with score=%f\n",
213  inl->frame_count_out, TS2D(in->pts) * av_q2d(inlink->time_base),
214  best_x, best_y, best_score);
215  foc->last_x = best_x;
216  foc->last_y = best_y;
217 
218  snprintf(buf, sizeof(buf), "%f", best_score);
219 
220  av_dict_set_int(&in->metadata, "lavfi.rect.w", foc->obj_frame->width, 0);
221  av_dict_set_int(&in->metadata, "lavfi.rect.h", foc->obj_frame->height, 0);
222  av_dict_set_int(&in->metadata, "lavfi.rect.x", best_x, 0);
223  av_dict_set_int(&in->metadata, "lavfi.rect.y", best_y, 0);
224  av_dict_set(&in->metadata, "lavfi.rect.score", buf, 0);
225 
226  return ff_filter_frame(ctx->outputs[0], in);
227 }
228 
230 {
231  FOCContext *foc = ctx->priv;
232  int i;
233 
234  for (i = 0; i < MAX_MIPMAPS; i++) {
235  av_frame_free(&foc->needle_frame[i]);
237  }
238 
239  if (foc->obj_frame)
240  av_freep(&foc->obj_frame->data[0]);
241  av_frame_free(&foc->obj_frame);
242 }
243 
245 {
246  FOCContext *foc = ctx->priv;
247  int ret, i;
248 
249  if (!foc->obj_filename) {
250  av_log(ctx, AV_LOG_ERROR, "object filename not set\n");
251  return AVERROR(EINVAL);
252  }
253 
254  foc->obj_frame = av_frame_alloc();
255  if (!foc->obj_frame)
256  return AVERROR(ENOMEM);
257 
258  if ((ret = ff_load_image(foc->obj_frame->data, foc->obj_frame->linesize,
259  &foc->obj_frame->width, &foc->obj_frame->height,
260  &foc->obj_frame->format, foc->obj_filename, ctx)) < 0)
261  return ret;
262 
263  if (foc->obj_frame->format != AV_PIX_FMT_GRAY8) {
264  av_log(ctx, AV_LOG_ERROR, "object image is not a grayscale image\n");
265  return AVERROR(EINVAL);
266  }
267 
268  foc->needle_frame[0] = av_frame_clone(foc->obj_frame);
269  for (i = 1; i < foc->mipmaps; i++) {
270  foc->needle_frame[i] = downscale(foc->needle_frame[i-1]);
271  if (!foc->needle_frame[i])
272  return AVERROR(ENOMEM);
273  }
274 
275  return 0;
276 }
277 
278 static const AVFilterPad foc_inputs[] = {
279  {
280  .name = "default",
281  .type = AVMEDIA_TYPE_VIDEO,
282  .config_props = config_input,
283  .filter_frame = filter_frame,
284  },
285 };
286 
288  .name = "find_rect",
289  .description = NULL_IF_CONFIG_SMALL("Find a user specified object."),
290  .priv_size = sizeof(FOCContext),
291  .init = init,
292  .uninit = uninit,
297  .priv_class = &find_rect_class,
298 };
FOCContext::threshold
float threshold
Definition: vf_find_rect.c:37
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:49
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:304
ff_filter_frame
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
Definition: avfilter.c:1061
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:162
FILTER_INPUTS
#define FILTER_INPUTS(array)
Definition: filters.h:262
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:389
AVFrame::pts
int64_t pts
Presentation timestamp in time_base units (time when frame should be shown to user).
Definition: frame.h:501
AVFrame::width
int width
Definition: frame.h:461
downscale
static AVFrame * downscale(AVFrame *in)
Definition: vf_find_rect.c:64
AVOption
AVOption.
Definition: opt.h:429
FOCContext::obj_filename
char * obj_filename
Definition: vf_find_rect.c:40
FFMAX
#define FFMAX(a, b)
Definition: macros.h:47
AVFilter::name
const char * name
Filter name.
Definition: avfilter.h:205
video.h
FOCContext::last_x
int last_x
Definition: vf_find_rect.c:41
AVFrame::data
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
Definition: frame.h:410
FOCContext::xmin
int xmin
Definition: vf_find_rect.c:39
AVFilterPad
A filter pad used for either input or output.
Definition: filters.h:38
av_frame_alloc
AVFrame * av_frame_alloc(void)
Allocate an AVFrame and set its fields to default values.
Definition: frame.c:150
AV_LOG_ERROR
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:209
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:174
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:45
filters.h
ctx
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:609
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:39
FILTER_OUTPUTS
#define FILTER_OUTPUTS(array)
Definition: filters.h:263
AVClass
Describe the class of an AVClass context structure.
Definition: log.h:75
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:197
TS2D
#define TS2D(ts)
Definition: filters.h:277
init
static av_cold int init(AVFilterContext *ctx)
Definition: vf_find_rect.c:244
FOCContext::mipmaps
int mipmaps
Definition: vf_find_rect.c:38
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:278
FOCContext
Definition: vf_find_rect.c:35
dst
uint8_t ptrdiff_t const uint8_t ptrdiff_t int intptr_t intptr_t int int16_t * dst
Definition: dsp.h:83
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:476
FOCContext::haystack_frame
AVFrame * haystack_frame[MAX_MIPMAPS]
Definition: vf_find_rect.c:44
OFFSET
#define OFFSET(x)
Definition: vf_find_rect.c:48
ff_vf_find_rect
const AVFilter ff_vf_find_rect
Definition: vf_find_rect.c:287
AV_LOG_INFO
#define AV_LOG_INFO
Standard information.
Definition: log.h:220
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
config_input
static int config_input(AVFilterLink *inlink)
Definition: vf_find_rect.c:135
FOCContext::last_y
int last_y
Definition: vf_find_rect.c:41
FFMIN
#define FFMIN(a, b)
Definition: macros.h:49
AVFilterPad::name
const char * name
Pad name.
Definition: filters.h:44
FILTER_PIXFMTS
#define FILTER_PIXFMTS(...)
Definition: filters.h:248
FOCContext::xmax
int xmax
Definition: vf_find_rect.c:39
AVFilter
Filter definition.
Definition: avfilter.h:201
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:148
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:43
FOCContext::ymin
int ymin
Definition: vf_find_rect.c:39
AVFrame::height
int height
Definition: frame.h:461
compare
static float compare(const AVFrame *haystack, const AVFrame *obj, int offx, int offy)
Definition: vf_find_rect.c:97
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:707
uninit
static av_cold void uninit(AVFilterContext *ctx)
Definition: vf_find_rect.c:229
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:168
find_rect_options
static const AVOption find_rect_options[]
Definition: vf_find_rect.c:50
lavfutils.h
MAX_MIPMAPS
#define MAX_MIPMAPS
Definition: vf_find_rect.c:33
AVFilterContext
An instance of a filter.
Definition: avfilter.h:457
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
mem.h
AV_OPT_TYPE_BOOL
@ AV_OPT_TYPE_BOOL
Underlying C type is int.
Definition: opt.h:327
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
FOCContext::obj_frame
AVFrame * obj_frame
Definition: vf_find_rect.c:42
flags
#define flags(name, subs,...)
Definition: cbs_av1.c:482
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:434
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