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/avassert.h"
26 #include "libavutil/imgutils.h"
27 #include "libavutil/opt.h"
30 #include "lavfutils.h"
34 typedef struct FOCContext {
38 int xmin, ymin, xmax, ymax;
42 AVFrame *needle_frame[MAX_MIPMAPS];
43 AVFrame *haystack_frame[MAX_MIPMAPS];
46 #define OFFSET(x) offsetof(FOCContext, x)
47 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
48 static const AVOption foc_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 },
59 static const AVClass foc_class = {
60 .class_name = "find_rect",
61 .item_name = av_default_item_name,
62 .option = foc_options,
63 .version = LIBAVUTIL_VERSION_INT,
64 .category = AV_CLASS_CATEGORY_FILTER,
67 static int query_formats(AVFilterContext *ctx)
69 static const enum AVPixelFormat pix_fmts[] = {
75 return ff_set_common_formats(ctx, ff_make_format_list(pix_fmts));
78 static AVFrame *downscale(AVFrame *in)
81 AVFrame *frame = av_frame_alloc();
86 frame->format = in->format;
87 frame->width = (in->width + 1) / 2;
88 frame->height = (in->height+ 1) / 2;
90 if (av_frame_get_buffer(frame, 32) < 0) {
91 av_frame_free(&frame);
97 for(y = 0; y < frame->height; y++) {
98 for(x = 0; x < frame->width; x++) {
101 + src[2*x+0 + in->linesize[0]]
102 + src[2*x+1 + in->linesize[0]]
105 src += 2*in->linesize[0];
106 dst += frame->linesize[0];
111 static float compare(const AVFrame *haystack, const AVFrame *obj, int offx, int offy)
116 int64_t oo_sum_v = 0;
117 int64_t hh_sum_v = 0;
118 int64_t oh_sum_v = 0;
120 int n = obj->height * obj->width;
121 const uint8_t *odat = obj ->data[0];
122 const uint8_t *hdat = haystack->data[0] + offx + offy * haystack->linesize[0];
123 int64_t o_sigma, h_sigma;
125 for(y = 0; y < obj->height; y++) {
126 for(x = 0; x < obj->width; x++) {
131 oo_sum_v += o_v * o_v;
132 hh_sum_v += h_v * h_v;
133 oh_sum_v += o_v * h_v;
135 odat += obj->linesize[0];
136 hdat += haystack->linesize[0];
138 o_sigma = n*oo_sum_v - o_sum_v*(int64_t)o_sum_v;
139 h_sigma = n*hh_sum_v - h_sum_v*(int64_t)h_sum_v;
141 if (o_sigma == 0 || h_sigma == 0)
144 c = (n*oh_sum_v - o_sum_v*(int64_t)h_sum_v) / (sqrt(o_sigma)*sqrt(h_sigma));
149 static int config_input(AVFilterLink *inlink)
151 AVFilterContext *ctx = inlink->dst;
152 FOCContext *foc = ctx->priv;
155 foc->xmax = inlink->w - foc->obj_frame->width;
157 foc->ymax = inlink->h - foc->obj_frame->height;
162 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)
166 if (pass + 1 <= maxpass) {
168 search(foc, pass+1, maxpass, xmin>>1, (xmax+1)>>1, ymin>>1, (ymax+1)>>1, &sub_x, &sub_y, 1.0);
169 xmin = FFMAX(xmin, 2*sub_x - 4);
170 xmax = FFMIN(xmax, 2*sub_x + 4);
171 ymin = FFMAX(ymin, 2*sub_y - 4);
172 ymax = FFMIN(ymax, 2*sub_y + 4);
175 for (y = ymin; y <= ymax; y++) {
176 for (x = xmin; x <= xmax; x++) {
177 float score = compare(foc->haystack_frame[pass], foc->needle_frame[pass], x, y);
178 av_assert0(score != 0);
179 if (score < best_score) {
189 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
191 AVFilterContext *ctx = inlink->dst;
192 FOCContext *foc = ctx->priv;
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]);
202 best_score = search(foc, 0, 0,
203 FFMAX(foc->xmin, foc->last_x - 8),
204 FFMIN(foc->xmax, foc->last_x + 8),
205 FFMAX(foc->ymin, foc->last_y - 8),
206 FFMIN(foc->ymax, foc->last_y + 8),
207 &best_x, &best_y, 1.0);
209 best_score = search(foc, 0, foc->mipmaps - 1, foc->xmin, foc->xmax, foc->ymin, foc->ymax,
210 &best_x, &best_y, best_score);
212 for (i=0; i<MAX_MIPMAPS; i++) {
213 av_frame_free(&foc->haystack_frame[i]);
216 if (best_score > foc->threshold) {
217 return ff_filter_frame(ctx->outputs[0], in);
220 av_log(ctx, AV_LOG_DEBUG, "Found at %d %d score %f\n", best_x, best_y, best_score);
221 foc->last_x = best_x;
222 foc->last_y = best_y;
224 av_frame_make_writable(in);
226 av_dict_set_int(&in->metadata, "lavfi.rect.w", foc->obj_frame->width, 0);
227 av_dict_set_int(&in->metadata, "lavfi.rect.h", foc->obj_frame->height, 0);
228 av_dict_set_int(&in->metadata, "lavfi.rect.x", best_x, 0);
229 av_dict_set_int(&in->metadata, "lavfi.rect.y", best_y, 0);
231 return ff_filter_frame(ctx->outputs[0], in);
234 static av_cold void uninit(AVFilterContext *ctx)
236 FOCContext *foc = ctx->priv;
239 for (i = 0; i < MAX_MIPMAPS; i++) {
240 av_frame_free(&foc->needle_frame[i]);
241 av_frame_free(&foc->haystack_frame[i]);
245 av_freep(&foc->obj_frame->data[0]);
246 av_frame_free(&foc->obj_frame);
249 static av_cold int init(AVFilterContext *ctx)
251 FOCContext *foc = ctx->priv;
254 if (!foc->obj_filename) {
255 av_log(ctx, AV_LOG_ERROR, "object filename not set\n");
256 return AVERROR(EINVAL);
259 foc->obj_frame = av_frame_alloc();
261 return AVERROR(ENOMEM);
263 if ((ret = ff_load_image(foc->obj_frame->data, foc->obj_frame->linesize,
264 &foc->obj_frame->width, &foc->obj_frame->height,
265 &foc->obj_frame->format, foc->obj_filename, ctx)) < 0)
268 if (foc->obj_frame->format != AV_PIX_FMT_GRAY8) {
269 av_log(ctx, AV_LOG_ERROR, "object image is not a grayscale image\n");
270 return AVERROR(EINVAL);
273 foc->needle_frame[0] = av_frame_clone(foc->obj_frame);
274 for (i = 1; i < foc->mipmaps; i++) {
275 foc->needle_frame[i] = downscale(foc->needle_frame[i-1]);
276 if (!foc->needle_frame[i])
277 return AVERROR(ENOMEM);
283 static const AVFilterPad foc_inputs[] = {
286 .type = AVMEDIA_TYPE_VIDEO,
287 .config_props = config_input,
288 .filter_frame = filter_frame,
293 static const AVFilterPad foc_outputs[] = {
296 .type = AVMEDIA_TYPE_VIDEO,
301 AVFilter ff_vf_find_rect = {
303 .description = NULL_IF_CONFIG_SMALL("Find a user specified object"),
304 .priv_size = sizeof(FOCContext),
307 .query_formats = query_formats,
308 .inputs = foc_inputs,
309 .outputs = foc_outputs,
310 .priv_class = &foc_class,