yading@10: /*
yading@10:  * Copyright (c) 2010 Stefano Sabatini
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: /**
yading@10:  * @file
yading@10:  * libopencv wrapper functions
yading@10:  */
yading@10: 
yading@10: /* #define DEBUG */
yading@10: 
yading@10: #include <opencv/cv.h>
yading@10: #include <opencv/cxcore.h>
yading@10: #include "libavutil/avstring.h"
yading@10: #include "libavutil/common.h"
yading@10: #include "libavutil/file.h"
yading@10: #include "libavutil/opt.h"
yading@10: #include "avfilter.h"
yading@10: #include "formats.h"
yading@10: #include "internal.h"
yading@10: #include "video.h"
yading@10: 
yading@10: static void fill_iplimage_from_frame(IplImage *img, const AVFrame *frame, enum AVPixelFormat pixfmt)
yading@10: {
yading@10:     IplImage *tmpimg;
yading@10:     int depth, channels_nb;
yading@10: 
yading@10:     if      (pixfmt == AV_PIX_FMT_GRAY8) { depth = IPL_DEPTH_8U;  channels_nb = 1; }
yading@10:     else if (pixfmt == AV_PIX_FMT_BGRA)  { depth = IPL_DEPTH_8U;  channels_nb = 4; }
yading@10:     else if (pixfmt == AV_PIX_FMT_BGR24) { depth = IPL_DEPTH_8U;  channels_nb = 3; }
yading@10:     else return;
yading@10: 
yading@10:     tmpimg = cvCreateImageHeader((CvSize){frame->width, frame->height}, depth, channels_nb);
yading@10:     *img = *tmpimg;
yading@10:     img->imageData = img->imageDataOrigin = frame->data[0];
yading@10:     img->dataOrder = IPL_DATA_ORDER_PIXEL;
yading@10:     img->origin    = IPL_ORIGIN_TL;
yading@10:     img->widthStep = frame->linesize[0];
yading@10: }
yading@10: 
yading@10: static void fill_frame_from_iplimage(AVFrame *frame, const IplImage *img, enum AVPixelFormat pixfmt)
yading@10: {
yading@10:     frame->linesize[0] = img->widthStep;
yading@10:     frame->data[0]     = img->imageData;
yading@10: }
yading@10: 
yading@10: static int query_formats(AVFilterContext *ctx)
yading@10: {
yading@10:     static const enum AVPixelFormat pix_fmts[] = {
yading@10:         AV_PIX_FMT_BGR24, AV_PIX_FMT_BGRA, AV_PIX_FMT_GRAY8, AV_PIX_FMT_NONE
yading@10:     };
yading@10: 
yading@10:     ff_set_common_formats(ctx, ff_make_format_list(pix_fmts));
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: typedef struct {
yading@10:     const AVClass *class;
yading@10:     char *name;
yading@10:     char *params;
yading@10:     int (*init)(AVFilterContext *ctx, const char *args);
yading@10:     void (*uninit)(AVFilterContext *ctx);
yading@10:     void (*end_frame_filter)(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg);
yading@10:     void *priv;
yading@10: } OCVContext;
yading@10: 
yading@10: typedef struct {
yading@10:     int type;
yading@10:     int    param1, param2;
yading@10:     double param3, param4;
yading@10: } SmoothContext;
yading@10: 
yading@10: static av_cold int smooth_init(AVFilterContext *ctx, const char *args)
yading@10: {
yading@10:     OCVContext *ocv = ctx->priv;
yading@10:     SmoothContext *smooth = ocv->priv;
yading@10:     char type_str[128] = "gaussian";
yading@10: 
yading@10:     smooth->param1 = 3;
yading@10:     smooth->param2 = 0;
yading@10:     smooth->param3 = 0.0;
yading@10:     smooth->param4 = 0.0;
yading@10: 
yading@10:     if (args)
yading@10:         sscanf(args, "%127[^|]|%d|%d|%lf|%lf", type_str, &smooth->param1, &smooth->param2, &smooth->param3, &smooth->param4);
yading@10: 
yading@10:     if      (!strcmp(type_str, "blur"         )) smooth->type = CV_BLUR;
yading@10:     else if (!strcmp(type_str, "blur_no_scale")) smooth->type = CV_BLUR_NO_SCALE;
yading@10:     else if (!strcmp(type_str, "median"       )) smooth->type = CV_MEDIAN;
yading@10:     else if (!strcmp(type_str, "gaussian"     )) smooth->type = CV_GAUSSIAN;
yading@10:     else if (!strcmp(type_str, "bilateral"    )) smooth->type = CV_BILATERAL;
yading@10:     else {
yading@10:         av_log(ctx, AV_LOG_ERROR, "Smoothing type '%s' unknown.\n", type_str);
yading@10:         return AVERROR(EINVAL);
yading@10:     }
yading@10: 
yading@10:     if (smooth->param1 < 0 || !(smooth->param1%2)) {
yading@10:         av_log(ctx, AV_LOG_ERROR,
yading@10:                "Invalid value '%d' for param1, it has to be a positive odd number\n",
yading@10:                smooth->param1);
yading@10:         return AVERROR(EINVAL);
yading@10:     }
yading@10:     if ((smooth->type == CV_BLUR || smooth->type == CV_BLUR_NO_SCALE || smooth->type == CV_GAUSSIAN) &&
yading@10:         (smooth->param2 < 0 || (smooth->param2 && !(smooth->param2%2)))) {
yading@10:         av_log(ctx, AV_LOG_ERROR,
yading@10:                "Invalid value '%d' for param2, it has to be zero or a positive odd number\n",
yading@10:                smooth->param2);
yading@10:         return AVERROR(EINVAL);
yading@10:     }
yading@10: 
yading@10:     av_log(ctx, AV_LOG_VERBOSE, "type:%s param1:%d param2:%d param3:%f param4:%f\n",
yading@10:            type_str, smooth->param1, smooth->param2, smooth->param3, smooth->param4);
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static void smooth_end_frame_filter(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg)
yading@10: {
yading@10:     OCVContext *ocv = ctx->priv;
yading@10:     SmoothContext *smooth = ocv->priv;
yading@10:     cvSmooth(inimg, outimg, smooth->type, smooth->param1, smooth->param2, smooth->param3, smooth->param4);
yading@10: }
yading@10: 
yading@10: static int read_shape_from_file(int *cols, int *rows, int **values, const char *filename,
yading@10:                                 void *log_ctx)
yading@10: {
yading@10:     uint8_t *buf, *p, *pend;
yading@10:     size_t size;
yading@10:     int ret, i, j, w;
yading@10: 
yading@10:     if ((ret = av_file_map(filename, &buf, &size, 0, log_ctx)) < 0)
yading@10:         return ret;
yading@10: 
yading@10:     /* prescan file to get the number of lines and the maximum width */
yading@10:     w = 0;
yading@10:     for (i = 0; i < size; i++) {
yading@10:         if (buf[i] == '\n') {
yading@10:             if (*rows == INT_MAX) {
yading@10:                 av_log(log_ctx, AV_LOG_ERROR, "Overflow on the number of rows in the file\n");
yading@10:                 return AVERROR_INVALIDDATA;
yading@10:             }
yading@10:             ++(*rows);
yading@10:             *cols = FFMAX(*cols, w);
yading@10:             w = 0;
yading@10:         } else if (w == INT_MAX) {
yading@10:             av_log(log_ctx, AV_LOG_ERROR, "Overflow on the number of columns in the file\n");
yading@10:             return AVERROR_INVALIDDATA;
yading@10:         }
yading@10:         w++;
yading@10:     }
yading@10:     if (*rows > (SIZE_MAX / sizeof(int) / *cols)) {
yading@10:         av_log(log_ctx, AV_LOG_ERROR, "File with size %dx%d is too big\n",
yading@10:                *rows, *cols);
yading@10:         return AVERROR_INVALIDDATA;
yading@10:     }
yading@10:     if (!(*values = av_mallocz(sizeof(int) * *rows * *cols)))
yading@10:         return AVERROR(ENOMEM);
yading@10: 
yading@10:     /* fill *values */
yading@10:     p    = buf;
yading@10:     pend = buf + size-1;
yading@10:     for (i = 0; i < *rows; i++) {
yading@10:         for (j = 0;; j++) {
yading@10:             if (p > pend || *p == '\n') {
yading@10:                 p++;
yading@10:                 break;
yading@10:             } else
yading@10:                 (*values)[*cols*i + j] = !!av_isgraph(*(p++));
yading@10:         }
yading@10:     }
yading@10:     av_file_unmap(buf, size);
yading@10: 
yading@10: #ifdef DEBUG
yading@10:     {
yading@10:         char *line;
yading@10:         if (!(line = av_malloc(*cols + 1)))
yading@10:             return AVERROR(ENOMEM);
yading@10:         for (i = 0; i < *rows; i++) {
yading@10:             for (j = 0; j < *cols; j++)
yading@10:                 line[j] = (*values)[i * *cols + j] ? '@' : ' ';
yading@10:             line[j] = 0;
yading@10:             av_log(log_ctx, AV_LOG_DEBUG, "%3d: %s\n", i, line);
yading@10:         }
yading@10:         av_free(line);
yading@10:     }
yading@10: #endif
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int parse_iplconvkernel(IplConvKernel **kernel, char *buf, void *log_ctx)
yading@10: {
yading@10:     char shape_filename[128] = "", shape_str[32] = "rect";
yading@10:     int cols = 0, rows = 0, anchor_x = 0, anchor_y = 0, shape = CV_SHAPE_RECT;
yading@10:     int *values = NULL, ret;
yading@10: 
yading@10:     sscanf(buf, "%dx%d+%dx%d/%32[^=]=%127s", &cols, &rows, &anchor_x, &anchor_y, shape_str, shape_filename);
yading@10: 
yading@10:     if      (!strcmp(shape_str, "rect"   )) shape = CV_SHAPE_RECT;
yading@10:     else if (!strcmp(shape_str, "cross"  )) shape = CV_SHAPE_CROSS;
yading@10:     else if (!strcmp(shape_str, "ellipse")) shape = CV_SHAPE_ELLIPSE;
yading@10:     else if (!strcmp(shape_str, "custom" )) {
yading@10:         shape = CV_SHAPE_CUSTOM;
yading@10:         if ((ret = read_shape_from_file(&cols, &rows, &values, shape_filename, log_ctx)) < 0)
yading@10:             return ret;
yading@10:     } else {
yading@10:         av_log(log_ctx, AV_LOG_ERROR,
yading@10:                "Shape unspecified or type '%s' unknown.\n", shape_str);
yading@10:         return AVERROR(EINVAL);
yading@10:     }
yading@10: 
yading@10:     if (rows <= 0 || cols <= 0) {
yading@10:         av_log(log_ctx, AV_LOG_ERROR,
yading@10:                "Invalid non-positive values for shape size %dx%d\n", cols, rows);
yading@10:         return AVERROR(EINVAL);
yading@10:     }
yading@10: 
yading@10:     if (anchor_x < 0 || anchor_y < 0 || anchor_x >= cols || anchor_y >= rows) {
yading@10:         av_log(log_ctx, AV_LOG_ERROR,
yading@10:                "Shape anchor %dx%d is not inside the rectangle with size %dx%d.\n",
yading@10:                anchor_x, anchor_y, cols, rows);
yading@10:         return AVERROR(EINVAL);
yading@10:     }
yading@10: 
yading@10:     *kernel = cvCreateStructuringElementEx(cols, rows, anchor_x, anchor_y, shape, values);
yading@10:     av_freep(&values);
yading@10:     if (!*kernel)
yading@10:         return AVERROR(ENOMEM);
yading@10: 
yading@10:     av_log(log_ctx, AV_LOG_VERBOSE, "Structuring element: w:%d h:%d x:%d y:%d shape:%s\n",
yading@10:            rows, cols, anchor_x, anchor_y, shape_str);
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: typedef struct {
yading@10:     int nb_iterations;
yading@10:     IplConvKernel *kernel;
yading@10: } DilateContext;
yading@10: 
yading@10: static av_cold int dilate_init(AVFilterContext *ctx, const char *args)
yading@10: {
yading@10:     OCVContext *ocv = ctx->priv;
yading@10:     DilateContext *dilate = ocv->priv;
yading@10:     char default_kernel_str[] = "3x3+0x0/rect";
yading@10:     char *kernel_str;
yading@10:     const char *buf = args;
yading@10:     int ret;
yading@10: 
yading@10:     if (args)
yading@10:         kernel_str = av_get_token(&buf, "|");
yading@10:     else
yading@10:         kernel_str = av_strdup(default_kernel_str);
yading@10:     if (!kernel_str)
yading@10:         return AVERROR(ENOMEM);
yading@10:     if ((ret = parse_iplconvkernel(&dilate->kernel, kernel_str, ctx)) < 0)
yading@10:         return ret;
yading@10:     av_free(kernel_str);
yading@10: 
yading@10:     if (!buf || sscanf(buf, "|%d", &dilate->nb_iterations) != 1)
yading@10:         dilate->nb_iterations = 1;
yading@10:     av_log(ctx, AV_LOG_VERBOSE, "iterations_nb:%d\n", dilate->nb_iterations);
yading@10:     if (dilate->nb_iterations <= 0) {
yading@10:         av_log(ctx, AV_LOG_ERROR, "Invalid non-positive value '%d' for nb_iterations\n",
yading@10:                dilate->nb_iterations);
yading@10:         return AVERROR(EINVAL);
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static av_cold void dilate_uninit(AVFilterContext *ctx)
yading@10: {
yading@10:     OCVContext *ocv = ctx->priv;
yading@10:     DilateContext *dilate = ocv->priv;
yading@10: 
yading@10:     cvReleaseStructuringElement(&dilate->kernel);
yading@10: }
yading@10: 
yading@10: static void dilate_end_frame_filter(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg)
yading@10: {
yading@10:     OCVContext *ocv = ctx->priv;
yading@10:     DilateContext *dilate = ocv->priv;
yading@10:     cvDilate(inimg, outimg, dilate->kernel, dilate->nb_iterations);
yading@10: }
yading@10: 
yading@10: static void erode_end_frame_filter(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg)
yading@10: {
yading@10:     OCVContext *ocv = ctx->priv;
yading@10:     DilateContext *dilate = ocv->priv;
yading@10:     cvErode(inimg, outimg, dilate->kernel, dilate->nb_iterations);
yading@10: }
yading@10: 
yading@10: typedef struct {
yading@10:     const char *name;
yading@10:     size_t priv_size;
yading@10:     int  (*init)(AVFilterContext *ctx, const char *args);
yading@10:     void (*uninit)(AVFilterContext *ctx);
yading@10:     void (*end_frame_filter)(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg);
yading@10: } OCVFilterEntry;
yading@10: 
yading@10: static OCVFilterEntry ocv_filter_entries[] = {
yading@10:     { "dilate", sizeof(DilateContext), dilate_init, dilate_uninit, dilate_end_frame_filter },
yading@10:     { "erode",  sizeof(DilateContext), dilate_init, dilate_uninit, erode_end_frame_filter  },
yading@10:     { "smooth", sizeof(SmoothContext), smooth_init, NULL, smooth_end_frame_filter },
yading@10: };
yading@10: 
yading@10: static av_cold int init(AVFilterContext *ctx)
yading@10: {
yading@10:     OCVContext *ocv = ctx->priv;
yading@10:     int i;
yading@10: 
yading@10:     if (!ocv->name) {
yading@10:         av_log(ctx, AV_LOG_ERROR, "No libopencv filter name specified\n");
yading@10:         return AVERROR(EINVAL);
yading@10:     }
yading@10:     for (i = 0; i < FF_ARRAY_ELEMS(ocv_filter_entries); i++) {
yading@10:         OCVFilterEntry *entry = &ocv_filter_entries[i];
yading@10:         if (!strcmp(ocv->name, entry->name)) {
yading@10:             ocv->init             = entry->init;
yading@10:             ocv->uninit           = entry->uninit;
yading@10:             ocv->end_frame_filter = entry->end_frame_filter;
yading@10: 
yading@10:             if (!(ocv->priv = av_mallocz(entry->priv_size)))
yading@10:                 return AVERROR(ENOMEM);
yading@10:             return ocv->init(ctx, ocv->params);
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     av_log(ctx, AV_LOG_ERROR, "No libopencv filter named '%s'\n", ocv->name);
yading@10:     return AVERROR(EINVAL);
yading@10: }
yading@10: 
yading@10: static av_cold void uninit(AVFilterContext *ctx)
yading@10: {
yading@10:     OCVContext *ocv = ctx->priv;
yading@10: 
yading@10:     if (ocv->uninit)
yading@10:         ocv->uninit(ctx);
yading@10:     av_free(ocv->priv);
yading@10: }
yading@10: 
yading@10: static int filter_frame(AVFilterLink *inlink, AVFrame *in)
yading@10: {
yading@10:     AVFilterContext *ctx = inlink->dst;
yading@10:     OCVContext *ocv = ctx->priv;
yading@10:     AVFilterLink *outlink= inlink->dst->outputs[0];
yading@10:     AVFrame *out;
yading@10:     IplImage inimg, outimg;
yading@10: 
yading@10:     out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
yading@10:     if (!out) {
yading@10:         av_frame_free(&in);
yading@10:         return AVERROR(ENOMEM);
yading@10:     }
yading@10:     av_frame_copy_props(out, in);
yading@10: 
yading@10:     fill_iplimage_from_frame(&inimg , in , inlink->format);
yading@10:     fill_iplimage_from_frame(&outimg, out, inlink->format);
yading@10:     ocv->end_frame_filter(ctx, &inimg, &outimg);
yading@10:     fill_frame_from_iplimage(out, &outimg, inlink->format);
yading@10: 
yading@10:     av_frame_free(&in);
yading@10: 
yading@10:     return ff_filter_frame(outlink, out);
yading@10: }
yading@10: 
yading@10: #define OFFSET(x) offsetof(OCVContext, x)
yading@10: #define FLAGS AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_FILTERING_PARAM
yading@10: static const AVOption options[] = {
yading@10:     { "filter_name",   NULL, OFFSET(name),   AV_OPT_TYPE_STRING, .flags = FLAGS },
yading@10:     { "filter_params", NULL, OFFSET(params), AV_OPT_TYPE_STRING, .flags = FLAGS },
yading@10:     { NULL },
yading@10: };
yading@10: 
yading@10: static const AVClass ocv_class = {
yading@10:     .class_name = "ocv",
yading@10:     .item_name  = av_default_item_name,
yading@10:     .option     = options,
yading@10:     .version    = LIBAVUTIL_VERSION_INT,
yading@10: };
yading@10: 
yading@10: static const AVFilterPad avfilter_vf_ocv_inputs[] = {
yading@10:     {
yading@10:         .name       = "default",
yading@10:         .type       = AVMEDIA_TYPE_VIDEO,
yading@10:         .filter_frame = filter_frame,
yading@10:     },
yading@10:     { NULL }
yading@10: };
yading@10: 
yading@10: static const AVFilterPad avfilter_vf_ocv_outputs[] = {
yading@10:     {
yading@10:         .name = "default",
yading@10:         .type = AVMEDIA_TYPE_VIDEO,
yading@10:     },
yading@10:     { NULL }
yading@10: };
yading@10: 
yading@10: AVFilter avfilter_vf_ocv = {
yading@10:     .name        = "ocv",
yading@10:     .description = NULL_IF_CONFIG_SMALL("Apply transform using libopencv."),
yading@10: 
yading@10:     .priv_size = sizeof(OCVContext),
yading@10:     .priv_class = &ocv_class,
yading@10: 
yading@10:     .query_formats = query_formats,
yading@10:     .init = init,
yading@10:     .uninit = uninit,
yading@10: 
yading@10:     .inputs    = avfilter_vf_ocv_inputs,
yading@10: 
yading@10:     .outputs   = avfilter_vf_ocv_outputs,
yading@10: };