2 * Copyright (c) 2014-2015 Michael Niedermayer <michaelni@gmx.at>
4 * This file is part of FFmpeg.
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.
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.
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.
22 * @todo switch to dualinput
25 #include "libavutil/imgutils.h"
26 #include "libavutil/opt.h"
29 #include "lavfutils.h"
33 typedef struct FOCContext {
37 int xmin, ymin, xmax, ymax;
41 AVFrame *needle_frame[MAX_MIPMAPS];
42 AVFrame *haystack_frame[MAX_MIPMAPS];
45 #define OFFSET(x) offsetof(FOCContext, x)
46 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
47 static const AVOption find_rect_options[] = {
48 { "object", "object bitmap filename", OFFSET(obj_filename), AV_OPT_TYPE_STRING, {.str = NULL}, .flags = FLAGS },
49 { "threshold", "set threshold", OFFSET(threshold), AV_OPT_TYPE_FLOAT, {.dbl = 0.5}, 0, 1.0, FLAGS },
50 { "mipmaps", "set mipmaps", OFFSET(mipmaps), AV_OPT_TYPE_INT, {.i64 = 3}, 1, MAX_MIPMAPS, FLAGS },
51 { "xmin", "", OFFSET(xmin), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
52 { "ymin", "", OFFSET(ymin), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
53 { "xmax", "", OFFSET(xmax), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
54 { "ymax", "", OFFSET(ymax), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
58 AVFILTER_DEFINE_CLASS(find_rect);
60 static int query_formats(AVFilterContext *ctx)
62 static const enum AVPixelFormat pix_fmts[] = {
68 return ff_set_common_formats(ctx, ff_make_format_list(pix_fmts));
71 static AVFrame *downscale(AVFrame *in)
74 AVFrame *frame = av_frame_alloc();
79 frame->format = in->format;
80 frame->width = (in->width + 1) / 2;
81 frame->height = (in->height+ 1) / 2;
83 if (av_frame_get_buffer(frame, 32) < 0) {
84 av_frame_free(&frame);
90 for(y = 0; y < frame->height; y++) {
91 for(x = 0; x < frame->width; x++) {
94 + src[2*x+0 + in->linesize[0]]
95 + src[2*x+1 + in->linesize[0]]
98 src += 2*in->linesize[0];
99 dst += frame->linesize[0];
104 static float compare(const AVFrame *haystack, const AVFrame *obj, int offx, int offy)
109 int64_t oo_sum_v = 0;
110 int64_t hh_sum_v = 0;
111 int64_t oh_sum_v = 0;
113 int n = obj->height * obj->width;
114 const uint8_t *odat = obj ->data[0];
115 const uint8_t *hdat = haystack->data[0] + offx + offy * haystack->linesize[0];
116 int64_t o_sigma, h_sigma;
118 for(y = 0; y < obj->height; y++) {
119 for(x = 0; x < obj->width; x++) {
124 oo_sum_v += o_v * o_v;
125 hh_sum_v += h_v * h_v;
126 oh_sum_v += o_v * h_v;
128 odat += obj->linesize[0];
129 hdat += haystack->linesize[0];
131 o_sigma = n*oo_sum_v - o_sum_v*(int64_t)o_sum_v;
132 h_sigma = n*hh_sum_v - h_sum_v*(int64_t)h_sum_v;
134 if (o_sigma == 0 || h_sigma == 0)
137 c = (n*oh_sum_v - o_sum_v*(int64_t)h_sum_v) / (sqrt(o_sigma)*sqrt(h_sigma));
142 static int config_input(AVFilterLink *inlink)
144 AVFilterContext *ctx = inlink->dst;
145 FOCContext *foc = ctx->priv;
148 foc->xmax = inlink->w - foc->obj_frame->width;
150 foc->ymax = inlink->h - foc->obj_frame->height;
155 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)
159 if (pass + 1 <= maxpass) {
161 search(foc, pass+1, maxpass, xmin>>1, (xmax+1)>>1, ymin>>1, (ymax+1)>>1, &sub_x, &sub_y, 2.0);
162 xmin = FFMAX(xmin, 2*sub_x - 4);
163 xmax = FFMIN(xmax, 2*sub_x + 4);
164 ymin = FFMAX(ymin, 2*sub_y - 4);
165 ymax = FFMIN(ymax, 2*sub_y + 4);
168 for (y = ymin; y <= ymax; y++) {
169 for (x = xmin; x <= xmax; x++) {
170 float score = compare(foc->haystack_frame[pass], foc->needle_frame[pass], x, y);
171 if (score < best_score) {
181 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
183 AVFilterContext *ctx = inlink->dst;
184 FOCContext *foc = ctx->priv;
189 foc->haystack_frame[0] = av_frame_clone(in);
190 for (i=1; i<foc->mipmaps; i++) {
191 foc->haystack_frame[i] = downscale(foc->haystack_frame[i-1]);
194 best_score = search(foc, 0, 0,
195 FFMAX(foc->xmin, foc->last_x - 8),
196 FFMIN(foc->xmax, foc->last_x + 8),
197 FFMAX(foc->ymin, foc->last_y - 8),
198 FFMIN(foc->ymax, foc->last_y + 8),
199 &best_x, &best_y, 2.0);
201 best_score = search(foc, 0, foc->mipmaps - 1, foc->xmin, foc->xmax, foc->ymin, foc->ymax,
202 &best_x, &best_y, best_score);
204 for (i=0; i<MAX_MIPMAPS; i++) {
205 av_frame_free(&foc->haystack_frame[i]);
208 if (best_score > foc->threshold) {
209 return ff_filter_frame(ctx->outputs[0], in);
212 av_log(ctx, AV_LOG_DEBUG, "Found at %d %d score %f\n", best_x, best_y, best_score);
213 foc->last_x = best_x;
214 foc->last_y = best_y;
216 av_frame_make_writable(in);
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);
223 return ff_filter_frame(ctx->outputs[0], in);
226 static av_cold void uninit(AVFilterContext *ctx)
228 FOCContext *foc = ctx->priv;
231 for (i = 0; i < MAX_MIPMAPS; i++) {
232 av_frame_free(&foc->needle_frame[i]);
233 av_frame_free(&foc->haystack_frame[i]);
237 av_freep(&foc->obj_frame->data[0]);
238 av_frame_free(&foc->obj_frame);
241 static av_cold int init(AVFilterContext *ctx)
243 FOCContext *foc = ctx->priv;
246 if (!foc->obj_filename) {
247 av_log(ctx, AV_LOG_ERROR, "object filename not set\n");
248 return AVERROR(EINVAL);
251 foc->obj_frame = av_frame_alloc();
253 return AVERROR(ENOMEM);
255 if ((ret = ff_load_image(foc->obj_frame->data, foc->obj_frame->linesize,
256 &foc->obj_frame->width, &foc->obj_frame->height,
257 &foc->obj_frame->format, foc->obj_filename, ctx)) < 0)
260 if (foc->obj_frame->format != AV_PIX_FMT_GRAY8) {
261 av_log(ctx, AV_LOG_ERROR, "object image is not a grayscale image\n");
262 return AVERROR(EINVAL);
265 foc->needle_frame[0] = av_frame_clone(foc->obj_frame);
266 for (i = 1; i < foc->mipmaps; i++) {
267 foc->needle_frame[i] = downscale(foc->needle_frame[i-1]);
268 if (!foc->needle_frame[i])
269 return AVERROR(ENOMEM);
275 static const AVFilterPad foc_inputs[] = {
278 .type = AVMEDIA_TYPE_VIDEO,
279 .config_props = config_input,
280 .filter_frame = filter_frame,
285 static const AVFilterPad foc_outputs[] = {
288 .type = AVMEDIA_TYPE_VIDEO,
293 AVFilter ff_vf_find_rect = {
295 .description = NULL_IF_CONFIG_SMALL("Find a user specified object."),
296 .priv_size = sizeof(FOCContext),
299 .query_formats = query_formats,
300 .inputs = foc_inputs,
301 .outputs = foc_outputs,
302 .priv_class = &find_rect_class,