2 * Copyright (c) 2010 Stefano Sabatini
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (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 GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 * libopencv wrapper functions
27 #if HAVE_OPENCV2_CORE_CORE_C_H
28 #include <opencv2/core/core_c.h>
29 #include <opencv2/imgproc/imgproc_c.h>
31 #include <opencv/cv.h>
32 #include <opencv/cxcore.h>
34 #include "libavutil/avstring.h"
35 #include "libavutil/common.h"
36 #include "libavutil/file.h"
37 #include "libavutil/opt.h"
43 static void fill_iplimage_from_frame(IplImage *img, const AVFrame *frame, enum AVPixelFormat pixfmt)
46 int depth, channels_nb;
48 if (pixfmt == AV_PIX_FMT_GRAY8) { depth = IPL_DEPTH_8U; channels_nb = 1; }
49 else if (pixfmt == AV_PIX_FMT_BGRA) { depth = IPL_DEPTH_8U; channels_nb = 4; }
50 else if (pixfmt == AV_PIX_FMT_BGR24) { depth = IPL_DEPTH_8U; channels_nb = 3; }
53 tmpimg = cvCreateImageHeader((CvSize){frame->width, frame->height}, depth, channels_nb);
55 img->imageData = img->imageDataOrigin = frame->data[0];
56 img->dataOrder = IPL_DATA_ORDER_PIXEL;
57 img->origin = IPL_ORIGIN_TL;
58 img->widthStep = frame->linesize[0];
61 static void fill_frame_from_iplimage(AVFrame *frame, const IplImage *img, enum AVPixelFormat pixfmt)
63 frame->linesize[0] = img->widthStep;
64 frame->data[0] = img->imageData;
67 static int query_formats(AVFilterContext *ctx)
69 static const enum AVPixelFormat pix_fmts[] = {
70 AV_PIX_FMT_BGR24, AV_PIX_FMT_BGRA, AV_PIX_FMT_GRAY8, AV_PIX_FMT_NONE
72 AVFilterFormats *fmts_list = ff_make_format_list(pix_fmts);
74 return AVERROR(ENOMEM);
75 return ff_set_common_formats(ctx, fmts_list);
78 typedef struct OCVContext {
82 int (*init)(AVFilterContext *ctx, const char *args);
83 void (*uninit)(AVFilterContext *ctx);
84 void (*end_frame_filter)(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg);
88 typedef struct SmoothContext {
91 double param3, param4;
94 static av_cold int smooth_init(AVFilterContext *ctx, const char *args)
96 OCVContext *s = ctx->priv;
97 SmoothContext *smooth = s->priv;
98 char type_str[128] = "gaussian";
102 smooth->param3 = 0.0;
103 smooth->param4 = 0.0;
106 sscanf(args, "%127[^|]|%d|%d|%lf|%lf", type_str, &smooth->param1, &smooth->param2, &smooth->param3, &smooth->param4);
108 if (!strcmp(type_str, "blur" )) smooth->type = CV_BLUR;
109 else if (!strcmp(type_str, "blur_no_scale")) smooth->type = CV_BLUR_NO_SCALE;
110 else if (!strcmp(type_str, "median" )) smooth->type = CV_MEDIAN;
111 else if (!strcmp(type_str, "gaussian" )) smooth->type = CV_GAUSSIAN;
112 else if (!strcmp(type_str, "bilateral" )) smooth->type = CV_BILATERAL;
114 av_log(ctx, AV_LOG_ERROR, "Smoothing type '%s' unknown.\n", type_str);
115 return AVERROR(EINVAL);
118 if (smooth->param1 < 0 || !(smooth->param1%2)) {
119 av_log(ctx, AV_LOG_ERROR,
120 "Invalid value '%d' for param1, it has to be a positive odd number\n",
122 return AVERROR(EINVAL);
124 if ((smooth->type == CV_BLUR || smooth->type == CV_BLUR_NO_SCALE || smooth->type == CV_GAUSSIAN) &&
125 (smooth->param2 < 0 || (smooth->param2 && !(smooth->param2%2)))) {
126 av_log(ctx, AV_LOG_ERROR,
127 "Invalid value '%d' for param2, it has to be zero or a positive odd number\n",
129 return AVERROR(EINVAL);
132 av_log(ctx, AV_LOG_VERBOSE, "type:%s param1:%d param2:%d param3:%f param4:%f\n",
133 type_str, smooth->param1, smooth->param2, smooth->param3, smooth->param4);
137 static void smooth_end_frame_filter(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg)
139 OCVContext *s = ctx->priv;
140 SmoothContext *smooth = s->priv;
141 cvSmooth(inimg, outimg, smooth->type, smooth->param1, smooth->param2, smooth->param3, smooth->param4);
144 static int read_shape_from_file(int *cols, int *rows, int **values, const char *filename,
147 uint8_t *buf, *p, *pend;
151 if ((ret = av_file_map(filename, &buf, &size, 0, log_ctx)) < 0)
154 /* prescan file to get the number of lines and the maximum width */
156 for (i = 0; i < size; i++) {
157 if (buf[i] == '\n') {
158 if (*rows == INT_MAX) {
159 av_log(log_ctx, AV_LOG_ERROR, "Overflow on the number of rows in the file\n");
160 return AVERROR_INVALIDDATA;
163 *cols = FFMAX(*cols, w);
165 } else if (w == INT_MAX) {
166 av_log(log_ctx, AV_LOG_ERROR, "Overflow on the number of columns in the file\n");
167 return AVERROR_INVALIDDATA;
171 if (*rows > (SIZE_MAX / sizeof(int) / *cols)) {
172 av_log(log_ctx, AV_LOG_ERROR, "File with size %dx%d is too big\n",
174 return AVERROR_INVALIDDATA;
176 if (!(*values = av_mallocz_array(sizeof(int) * *rows, *cols)))
177 return AVERROR(ENOMEM);
182 for (i = 0; i < *rows; i++) {
184 if (p > pend || *p == '\n') {
188 (*values)[*cols*i + j] = !!av_isgraph(*(p++));
191 av_file_unmap(buf, size);
196 if (!(line = av_malloc(*cols + 1)))
197 return AVERROR(ENOMEM);
198 for (i = 0; i < *rows; i++) {
199 for (j = 0; j < *cols; j++)
200 line[j] = (*values)[i * *cols + j] ? '@' : ' ';
202 av_log(log_ctx, AV_LOG_DEBUG, "%3d: %s\n", i, line);
211 static int parse_iplconvkernel(IplConvKernel **kernel, char *buf, void *log_ctx)
213 char shape_filename[128] = "", shape_str[32] = "rect";
214 int cols = 0, rows = 0, anchor_x = 0, anchor_y = 0, shape = CV_SHAPE_RECT;
215 int *values = NULL, ret = 0;
217 sscanf(buf, "%dx%d+%dx%d/%32[^=]=%127s", &cols, &rows, &anchor_x, &anchor_y, shape_str, shape_filename);
219 if (!strcmp(shape_str, "rect" )) shape = CV_SHAPE_RECT;
220 else if (!strcmp(shape_str, "cross" )) shape = CV_SHAPE_CROSS;
221 else if (!strcmp(shape_str, "ellipse")) shape = CV_SHAPE_ELLIPSE;
222 else if (!strcmp(shape_str, "custom" )) {
223 shape = CV_SHAPE_CUSTOM;
224 if ((ret = read_shape_from_file(&cols, &rows, &values, shape_filename, log_ctx)) < 0)
227 av_log(log_ctx, AV_LOG_ERROR,
228 "Shape unspecified or type '%s' unknown.\n", shape_str);
229 ret = AVERROR(EINVAL);
233 if (rows <= 0 || cols <= 0) {
234 av_log(log_ctx, AV_LOG_ERROR,
235 "Invalid non-positive values for shape size %dx%d\n", cols, rows);
236 ret = AVERROR(EINVAL);
240 if (anchor_x < 0 || anchor_y < 0 || anchor_x >= cols || anchor_y >= rows) {
241 av_log(log_ctx, AV_LOG_ERROR,
242 "Shape anchor %dx%d is not inside the rectangle with size %dx%d.\n",
243 anchor_x, anchor_y, cols, rows);
244 ret = AVERROR(EINVAL);
248 *kernel = cvCreateStructuringElementEx(cols, rows, anchor_x, anchor_y, shape, values);
250 ret = AVERROR(ENOMEM);
254 av_log(log_ctx, AV_LOG_VERBOSE, "Structuring element: w:%d h:%d x:%d y:%d shape:%s\n",
255 rows, cols, anchor_x, anchor_y, shape_str);
261 typedef struct DilateContext {
263 IplConvKernel *kernel;
266 static av_cold int dilate_init(AVFilterContext *ctx, const char *args)
268 OCVContext *s = ctx->priv;
269 DilateContext *dilate = s->priv;
270 char default_kernel_str[] = "3x3+0x0/rect";
271 char *kernel_str = NULL;
272 const char *buf = args;
276 kernel_str = av_get_token(&buf, "|");
279 return AVERROR(ENOMEM);
282 ret = parse_iplconvkernel(&dilate->kernel,
283 (!kernel_str || !*kernel_str) ? default_kernel_str
290 if (!buf || sscanf(buf, "|%d", &dilate->nb_iterations) != 1)
291 dilate->nb_iterations = 1;
292 av_log(ctx, AV_LOG_VERBOSE, "iterations_nb:%d\n", dilate->nb_iterations);
293 if (dilate->nb_iterations <= 0) {
294 av_log(ctx, AV_LOG_ERROR, "Invalid non-positive value '%d' for nb_iterations\n",
295 dilate->nb_iterations);
296 return AVERROR(EINVAL);
301 static av_cold void dilate_uninit(AVFilterContext *ctx)
303 OCVContext *s = ctx->priv;
304 DilateContext *dilate = s->priv;
306 cvReleaseStructuringElement(&dilate->kernel);
309 static void dilate_end_frame_filter(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg)
311 OCVContext *s = ctx->priv;
312 DilateContext *dilate = s->priv;
313 cvDilate(inimg, outimg, dilate->kernel, dilate->nb_iterations);
316 static void erode_end_frame_filter(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg)
318 OCVContext *s = ctx->priv;
319 DilateContext *dilate = s->priv;
320 cvErode(inimg, outimg, dilate->kernel, dilate->nb_iterations);
323 typedef struct OCVFilterEntry {
326 int (*init)(AVFilterContext *ctx, const char *args);
327 void (*uninit)(AVFilterContext *ctx);
328 void (*end_frame_filter)(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg);
331 static const OCVFilterEntry ocv_filter_entries[] = {
332 { "dilate", sizeof(DilateContext), dilate_init, dilate_uninit, dilate_end_frame_filter },
333 { "erode", sizeof(DilateContext), dilate_init, dilate_uninit, erode_end_frame_filter },
334 { "smooth", sizeof(SmoothContext), smooth_init, NULL, smooth_end_frame_filter },
337 static av_cold int init(AVFilterContext *ctx)
339 OCVContext *s = ctx->priv;
343 av_log(ctx, AV_LOG_ERROR, "No libopencv filter name specified\n");
344 return AVERROR(EINVAL);
346 for (i = 0; i < FF_ARRAY_ELEMS(ocv_filter_entries); i++) {
347 const OCVFilterEntry *entry = &ocv_filter_entries[i];
348 if (!strcmp(s->name, entry->name)) {
349 s->init = entry->init;
350 s->uninit = entry->uninit;
351 s->end_frame_filter = entry->end_frame_filter;
353 if (!(s->priv = av_mallocz(entry->priv_size)))
354 return AVERROR(ENOMEM);
355 return s->init(ctx, s->params);
359 av_log(ctx, AV_LOG_ERROR, "No libopencv filter named '%s'\n", s->name);
360 return AVERROR(EINVAL);
363 static av_cold void uninit(AVFilterContext *ctx)
365 OCVContext *s = ctx->priv;
372 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
374 AVFilterContext *ctx = inlink->dst;
375 OCVContext *s = ctx->priv;
376 AVFilterLink *outlink= inlink->dst->outputs[0];
378 IplImage inimg, outimg;
380 out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
383 return AVERROR(ENOMEM);
385 av_frame_copy_props(out, in);
387 fill_iplimage_from_frame(&inimg , in , inlink->format);
388 fill_iplimage_from_frame(&outimg, out, inlink->format);
389 s->end_frame_filter(ctx, &inimg, &outimg);
390 fill_frame_from_iplimage(out, &outimg, inlink->format);
394 return ff_filter_frame(outlink, out);
397 #define OFFSET(x) offsetof(OCVContext, x)
398 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_FILTERING_PARAM
399 static const AVOption ocv_options[] = {
400 { "filter_name", NULL, OFFSET(name), AV_OPT_TYPE_STRING, .flags = FLAGS },
401 { "filter_params", NULL, OFFSET(params), AV_OPT_TYPE_STRING, .flags = FLAGS },
405 AVFILTER_DEFINE_CLASS(ocv);
407 static const AVFilterPad avfilter_vf_ocv_inputs[] = {
410 .type = AVMEDIA_TYPE_VIDEO,
411 .filter_frame = filter_frame,
416 static const AVFilterPad avfilter_vf_ocv_outputs[] = {
419 .type = AVMEDIA_TYPE_VIDEO,
424 AVFilter ff_vf_ocv = {
426 .description = NULL_IF_CONFIG_SMALL("Apply transform using libopencv."),
427 .priv_size = sizeof(OCVContext),
428 .priv_class = &ocv_class,
429 .query_formats = query_formats,
432 .inputs = avfilter_vf_ocv_inputs,
433 .outputs = avfilter_vf_ocv_outputs,