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
26 #include <opencv/cv.h>
27 #include <opencv/cxcore.h>
28 #include "libavutil/avstring.h"
29 #include "libavutil/common.h"
30 #include "libavutil/file.h"
31 #include "libavutil/opt.h"
37 static void fill_iplimage_from_frame(IplImage *img, const AVFrame *frame, enum AVPixelFormat pixfmt)
40 int depth, channels_nb;
42 if (pixfmt == AV_PIX_FMT_GRAY8) { depth = IPL_DEPTH_8U; channels_nb = 1; }
43 else if (pixfmt == AV_PIX_FMT_BGRA) { depth = IPL_DEPTH_8U; channels_nb = 4; }
44 else if (pixfmt == AV_PIX_FMT_BGR24) { depth = IPL_DEPTH_8U; channels_nb = 3; }
47 tmpimg = cvCreateImageHeader((CvSize){frame->width, frame->height}, depth, channels_nb);
49 img->imageData = img->imageDataOrigin = frame->data[0];
50 img->dataOrder = IPL_DATA_ORDER_PIXEL;
51 img->origin = IPL_ORIGIN_TL;
52 img->widthStep = frame->linesize[0];
55 static void fill_frame_from_iplimage(AVFrame *frame, const IplImage *img, enum AVPixelFormat pixfmt)
57 frame->linesize[0] = img->widthStep;
58 frame->data[0] = img->imageData;
61 static int query_formats(AVFilterContext *ctx)
63 static const enum AVPixelFormat pix_fmts[] = {
64 AV_PIX_FMT_BGR24, AV_PIX_FMT_BGRA, AV_PIX_FMT_GRAY8, AV_PIX_FMT_NONE
66 AVFilterFormats *fmts_list = ff_make_format_list(pix_fmts);
68 return AVERROR(ENOMEM);
69 return ff_set_common_formats(ctx, fmts_list);
72 typedef struct OCVContext {
76 int (*init)(AVFilterContext *ctx, const char *args);
77 void (*uninit)(AVFilterContext *ctx);
78 void (*end_frame_filter)(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg);
82 typedef struct SmoothContext {
85 double param3, param4;
88 static av_cold int smooth_init(AVFilterContext *ctx, const char *args)
90 OCVContext *s = ctx->priv;
91 SmoothContext *smooth = s->priv;
92 char type_str[128] = "gaussian";
100 sscanf(args, "%127[^|]|%d|%d|%lf|%lf", type_str, &smooth->param1, &smooth->param2, &smooth->param3, &smooth->param4);
102 if (!strcmp(type_str, "blur" )) smooth->type = CV_BLUR;
103 else if (!strcmp(type_str, "blur_no_scale")) smooth->type = CV_BLUR_NO_SCALE;
104 else if (!strcmp(type_str, "median" )) smooth->type = CV_MEDIAN;
105 else if (!strcmp(type_str, "gaussian" )) smooth->type = CV_GAUSSIAN;
106 else if (!strcmp(type_str, "bilateral" )) smooth->type = CV_BILATERAL;
108 av_log(ctx, AV_LOG_ERROR, "Smoothing type '%s' unknown.\n", type_str);
109 return AVERROR(EINVAL);
112 if (smooth->param1 < 0 || !(smooth->param1%2)) {
113 av_log(ctx, AV_LOG_ERROR,
114 "Invalid value '%d' for param1, it has to be a positive odd number\n",
116 return AVERROR(EINVAL);
118 if ((smooth->type == CV_BLUR || smooth->type == CV_BLUR_NO_SCALE || smooth->type == CV_GAUSSIAN) &&
119 (smooth->param2 < 0 || (smooth->param2 && !(smooth->param2%2)))) {
120 av_log(ctx, AV_LOG_ERROR,
121 "Invalid value '%d' for param2, it has to be zero or a positive odd number\n",
123 return AVERROR(EINVAL);
126 av_log(ctx, AV_LOG_VERBOSE, "type:%s param1:%d param2:%d param3:%f param4:%f\n",
127 type_str, smooth->param1, smooth->param2, smooth->param3, smooth->param4);
131 static void smooth_end_frame_filter(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg)
133 OCVContext *s = ctx->priv;
134 SmoothContext *smooth = s->priv;
135 cvSmooth(inimg, outimg, smooth->type, smooth->param1, smooth->param2, smooth->param3, smooth->param4);
138 static int read_shape_from_file(int *cols, int *rows, int **values, const char *filename,
141 uint8_t *buf, *p, *pend;
145 if ((ret = av_file_map(filename, &buf, &size, 0, log_ctx)) < 0)
148 /* prescan file to get the number of lines and the maximum width */
150 for (i = 0; i < size; i++) {
151 if (buf[i] == '\n') {
152 if (*rows == INT_MAX) {
153 av_log(log_ctx, AV_LOG_ERROR, "Overflow on the number of rows in the file\n");
154 return AVERROR_INVALIDDATA;
157 *cols = FFMAX(*cols, w);
159 } else if (w == INT_MAX) {
160 av_log(log_ctx, AV_LOG_ERROR, "Overflow on the number of columns in the file\n");
161 return AVERROR_INVALIDDATA;
165 if (*rows > (SIZE_MAX / sizeof(int) / *cols)) {
166 av_log(log_ctx, AV_LOG_ERROR, "File with size %dx%d is too big\n",
168 return AVERROR_INVALIDDATA;
170 if (!(*values = av_mallocz_array(sizeof(int) * *rows, *cols)))
171 return AVERROR(ENOMEM);
176 for (i = 0; i < *rows; i++) {
178 if (p > pend || *p == '\n') {
182 (*values)[*cols*i + j] = !!av_isgraph(*(p++));
185 av_file_unmap(buf, size);
190 if (!(line = av_malloc(*cols + 1)))
191 return AVERROR(ENOMEM);
192 for (i = 0; i < *rows; i++) {
193 for (j = 0; j < *cols; j++)
194 line[j] = (*values)[i * *cols + j] ? '@' : ' ';
196 av_log(log_ctx, AV_LOG_DEBUG, "%3d: %s\n", i, line);
205 static int parse_iplconvkernel(IplConvKernel **kernel, char *buf, void *log_ctx)
207 char shape_filename[128] = "", shape_str[32] = "rect";
208 int cols = 0, rows = 0, anchor_x = 0, anchor_y = 0, shape = CV_SHAPE_RECT;
209 int *values = NULL, ret = 0;
211 sscanf(buf, "%dx%d+%dx%d/%32[^=]=%127s", &cols, &rows, &anchor_x, &anchor_y, shape_str, shape_filename);
213 if (!strcmp(shape_str, "rect" )) shape = CV_SHAPE_RECT;
214 else if (!strcmp(shape_str, "cross" )) shape = CV_SHAPE_CROSS;
215 else if (!strcmp(shape_str, "ellipse")) shape = CV_SHAPE_ELLIPSE;
216 else if (!strcmp(shape_str, "custom" )) {
217 shape = CV_SHAPE_CUSTOM;
218 if ((ret = read_shape_from_file(&cols, &rows, &values, shape_filename, log_ctx)) < 0)
221 av_log(log_ctx, AV_LOG_ERROR,
222 "Shape unspecified or type '%s' unknown.\n", shape_str);
223 ret = AVERROR(EINVAL);
227 if (rows <= 0 || cols <= 0) {
228 av_log(log_ctx, AV_LOG_ERROR,
229 "Invalid non-positive values for shape size %dx%d\n", cols, rows);
230 ret = AVERROR(EINVAL);
234 if (anchor_x < 0 || anchor_y < 0 || anchor_x >= cols || anchor_y >= rows) {
235 av_log(log_ctx, AV_LOG_ERROR,
236 "Shape anchor %dx%d is not inside the rectangle with size %dx%d.\n",
237 anchor_x, anchor_y, cols, rows);
238 ret = AVERROR(EINVAL);
242 *kernel = cvCreateStructuringElementEx(cols, rows, anchor_x, anchor_y, shape, values);
244 ret = AVERROR(ENOMEM);
248 av_log(log_ctx, AV_LOG_VERBOSE, "Structuring element: w:%d h:%d x:%d y:%d shape:%s\n",
249 rows, cols, anchor_x, anchor_y, shape_str);
255 typedef struct DilateContext {
257 IplConvKernel *kernel;
260 static av_cold int dilate_init(AVFilterContext *ctx, const char *args)
262 OCVContext *s = ctx->priv;
263 DilateContext *dilate = s->priv;
264 char default_kernel_str[] = "3x3+0x0/rect";
265 char *kernel_str = NULL;
266 const char *buf = args;
270 kernel_str = av_get_token(&buf, "|");
273 return AVERROR(ENOMEM);
276 ret = parse_iplconvkernel(&dilate->kernel,
277 (!kernel_str || !*kernel_str) ? default_kernel_str
284 if (!buf || sscanf(buf, "|%d", &dilate->nb_iterations) != 1)
285 dilate->nb_iterations = 1;
286 av_log(ctx, AV_LOG_VERBOSE, "iterations_nb:%d\n", dilate->nb_iterations);
287 if (dilate->nb_iterations <= 0) {
288 av_log(ctx, AV_LOG_ERROR, "Invalid non-positive value '%d' for nb_iterations\n",
289 dilate->nb_iterations);
290 return AVERROR(EINVAL);
295 static av_cold void dilate_uninit(AVFilterContext *ctx)
297 OCVContext *s = ctx->priv;
298 DilateContext *dilate = s->priv;
300 cvReleaseStructuringElement(&dilate->kernel);
303 static void dilate_end_frame_filter(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg)
305 OCVContext *s = ctx->priv;
306 DilateContext *dilate = s->priv;
307 cvDilate(inimg, outimg, dilate->kernel, dilate->nb_iterations);
310 static void erode_end_frame_filter(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg)
312 OCVContext *s = ctx->priv;
313 DilateContext *dilate = s->priv;
314 cvErode(inimg, outimg, dilate->kernel, dilate->nb_iterations);
317 typedef struct OCVFilterEntry {
320 int (*init)(AVFilterContext *ctx, const char *args);
321 void (*uninit)(AVFilterContext *ctx);
322 void (*end_frame_filter)(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg);
325 static const OCVFilterEntry ocv_filter_entries[] = {
326 { "dilate", sizeof(DilateContext), dilate_init, dilate_uninit, dilate_end_frame_filter },
327 { "erode", sizeof(DilateContext), dilate_init, dilate_uninit, erode_end_frame_filter },
328 { "smooth", sizeof(SmoothContext), smooth_init, NULL, smooth_end_frame_filter },
331 static av_cold int init(AVFilterContext *ctx)
333 OCVContext *s = ctx->priv;
337 av_log(ctx, AV_LOG_ERROR, "No libopencv filter name specified\n");
338 return AVERROR(EINVAL);
340 for (i = 0; i < FF_ARRAY_ELEMS(ocv_filter_entries); i++) {
341 const OCVFilterEntry *entry = &ocv_filter_entries[i];
342 if (!strcmp(s->name, entry->name)) {
343 s->init = entry->init;
344 s->uninit = entry->uninit;
345 s->end_frame_filter = entry->end_frame_filter;
347 if (!(s->priv = av_mallocz(entry->priv_size)))
348 return AVERROR(ENOMEM);
349 return s->init(ctx, s->params);
353 av_log(ctx, AV_LOG_ERROR, "No libopencv filter named '%s'\n", s->name);
354 return AVERROR(EINVAL);
357 static av_cold void uninit(AVFilterContext *ctx)
359 OCVContext *s = ctx->priv;
366 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
368 AVFilterContext *ctx = inlink->dst;
369 OCVContext *s = ctx->priv;
370 AVFilterLink *outlink= inlink->dst->outputs[0];
372 IplImage inimg, outimg;
374 out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
377 return AVERROR(ENOMEM);
379 av_frame_copy_props(out, in);
381 fill_iplimage_from_frame(&inimg , in , inlink->format);
382 fill_iplimage_from_frame(&outimg, out, inlink->format);
383 s->end_frame_filter(ctx, &inimg, &outimg);
384 fill_frame_from_iplimage(out, &outimg, inlink->format);
388 return ff_filter_frame(outlink, out);
391 #define OFFSET(x) offsetof(OCVContext, x)
392 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_FILTERING_PARAM
393 static const AVOption ocv_options[] = {
394 { "filter_name", NULL, OFFSET(name), AV_OPT_TYPE_STRING, .flags = FLAGS },
395 { "filter_params", NULL, OFFSET(params), AV_OPT_TYPE_STRING, .flags = FLAGS },
399 AVFILTER_DEFINE_CLASS(ocv);
401 static const AVFilterPad avfilter_vf_ocv_inputs[] = {
404 .type = AVMEDIA_TYPE_VIDEO,
405 .filter_frame = filter_frame,
410 static const AVFilterPad avfilter_vf_ocv_outputs[] = {
413 .type = AVMEDIA_TYPE_VIDEO,
418 AVFilter ff_vf_ocv = {
420 .description = NULL_IF_CONFIG_SMALL("Apply transform using libopencv."),
421 .priv_size = sizeof(OCVContext),
422 .priv_class = &ocv_class,
423 .query_formats = query_formats,
426 .inputs = avfilter_vf_ocv_inputs,
427 .outputs = avfilter_vf_ocv_outputs,