yading@10: /*
yading@10:  * Copyright (c) 2012 Clément Bœsch
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:  * Edge detection filter
yading@10:  *
yading@10:  * @see https://en.wikipedia.org/wiki/Canny_edge_detector
yading@10:  */
yading@10: 
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: typedef struct {
yading@10:     const AVClass *class;
yading@10:     uint8_t  *tmpbuf;
yading@10:     uint16_t *gradients;
yading@10:     char     *directions;
yading@10:     double   low, high;
yading@10:     uint8_t  low_u8, high_u8;
yading@10: } EdgeDetectContext;
yading@10: 
yading@10: #define OFFSET(x) offsetof(EdgeDetectContext, x)
yading@10: #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
yading@10: static const AVOption edgedetect_options[] = {
yading@10:     { "high", "set high threshold", OFFSET(high), AV_OPT_TYPE_DOUBLE, {.dbl=50/255.}, 0, 1, FLAGS },
yading@10:     { "low",  "set low threshold",  OFFSET(low),  AV_OPT_TYPE_DOUBLE, {.dbl=20/255.}, 0, 1, FLAGS },
yading@10:     { NULL },
yading@10: };
yading@10: 
yading@10: AVFILTER_DEFINE_CLASS(edgedetect);
yading@10: 
yading@10: static av_cold int init(AVFilterContext *ctx)
yading@10: {
yading@10:     EdgeDetectContext *edgedetect = ctx->priv;
yading@10: 
yading@10:     edgedetect->low_u8  = edgedetect->low  * 255. + .5;
yading@10:     edgedetect->high_u8 = edgedetect->high * 255. + .5;
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int query_formats(AVFilterContext *ctx)
yading@10: {
yading@10:     static const enum AVPixelFormat pix_fmts[] = {AV_PIX_FMT_GRAY8, AV_PIX_FMT_NONE};
yading@10:     ff_set_common_formats(ctx, ff_make_format_list(pix_fmts));
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int config_props(AVFilterLink *inlink)
yading@10: {
yading@10:     AVFilterContext *ctx = inlink->dst;
yading@10:     EdgeDetectContext *edgedetect = ctx->priv;
yading@10: 
yading@10:     edgedetect->tmpbuf     = av_malloc(inlink->w * inlink->h);
yading@10:     edgedetect->gradients  = av_calloc(inlink->w * inlink->h, sizeof(*edgedetect->gradients));
yading@10:     edgedetect->directions = av_malloc(inlink->w * inlink->h);
yading@10:     if (!edgedetect->tmpbuf || !edgedetect->gradients || !edgedetect->directions)
yading@10:         return AVERROR(ENOMEM);
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static void gaussian_blur(AVFilterContext *ctx, int w, int h,
yading@10:                                 uint8_t *dst, int dst_linesize,
yading@10:                           const uint8_t *src, int src_linesize)
yading@10: {
yading@10:     int i, j;
yading@10: 
yading@10:     memcpy(dst, src, w); dst += dst_linesize; src += src_linesize;
yading@10:     memcpy(dst, src, w); dst += dst_linesize; src += src_linesize;
yading@10:     for (j = 2; j < h - 2; j++) {
yading@10:         dst[0] = src[0];
yading@10:         dst[1] = src[1];
yading@10:         for (i = 2; i < w - 2; i++) {
yading@10:             /* Gaussian mask of size 5x5 with sigma = 1.4 */
yading@10:             dst[i] = ((src[-2*src_linesize + i-2] + src[2*src_linesize + i-2]) * 2
yading@10:                     + (src[-2*src_linesize + i-1] + src[2*src_linesize + i-1]) * 4
yading@10:                     + (src[-2*src_linesize + i  ] + src[2*src_linesize + i  ]) * 5
yading@10:                     + (src[-2*src_linesize + i+1] + src[2*src_linesize + i+1]) * 4
yading@10:                     + (src[-2*src_linesize + i+2] + src[2*src_linesize + i+2]) * 2
yading@10: 
yading@10:                     + (src[  -src_linesize + i-2] + src[  src_linesize + i-2]) *  4
yading@10:                     + (src[  -src_linesize + i-1] + src[  src_linesize + i-1]) *  9
yading@10:                     + (src[  -src_linesize + i  ] + src[  src_linesize + i  ]) * 12
yading@10:                     + (src[  -src_linesize + i+1] + src[  src_linesize + i+1]) *  9
yading@10:                     + (src[  -src_linesize + i+2] + src[  src_linesize + i+2]) *  4
yading@10: 
yading@10:                     + src[i-2] *  5
yading@10:                     + src[i-1] * 12
yading@10:                     + src[i  ] * 15
yading@10:                     + src[i+1] * 12
yading@10:                     + src[i+2] *  5) / 159;
yading@10:         }
yading@10:         dst[i    ] = src[i    ];
yading@10:         dst[i + 1] = src[i + 1];
yading@10: 
yading@10:         dst += dst_linesize;
yading@10:         src += src_linesize;
yading@10:     }
yading@10:     memcpy(dst, src, w); dst += dst_linesize; src += src_linesize;
yading@10:     memcpy(dst, src, w);
yading@10: }
yading@10: 
yading@10: enum {
yading@10:     DIRECTION_45UP,
yading@10:     DIRECTION_45DOWN,
yading@10:     DIRECTION_HORIZONTAL,
yading@10:     DIRECTION_VERTICAL,
yading@10: };
yading@10: 
yading@10: static int get_rounded_direction(int gx, int gy)
yading@10: {
yading@10:     /* reference angles:
yading@10:      *   tan( pi/8) = sqrt(2)-1
yading@10:      *   tan(3pi/8) = sqrt(2)+1
yading@10:      * Gy/Gx is the tangent of the angle (theta), so Gy/Gx is compared against
yading@10:      * <ref-angle>, or more simply Gy against <ref-angle>*Gx
yading@10:      *
yading@10:      * Gx and Gy bounds = [-1020;1020], using 16-bit arithmetic:
yading@10:      *   round((sqrt(2)-1) * (1<<16)) =  27146
yading@10:      *   round((sqrt(2)+1) * (1<<16)) = 158218
yading@10:      */
yading@10:     if (gx) {
yading@10:         int tanpi8gx, tan3pi8gx;
yading@10: 
yading@10:         if (gx < 0)
yading@10:             gx = -gx, gy = -gy;
yading@10:         gy <<= 16;
yading@10:         tanpi8gx  =  27146 * gx;
yading@10:         tan3pi8gx = 158218 * gx;
yading@10:         if (gy > -tan3pi8gx && gy < -tanpi8gx)  return DIRECTION_45UP;
yading@10:         if (gy > -tanpi8gx  && gy <  tanpi8gx)  return DIRECTION_HORIZONTAL;
yading@10:         if (gy >  tanpi8gx  && gy <  tan3pi8gx) return DIRECTION_45DOWN;
yading@10:     }
yading@10:     return DIRECTION_VERTICAL;
yading@10: }
yading@10: 
yading@10: static void sobel(AVFilterContext *ctx, int w, int h,
yading@10:                         uint16_t *dst, int dst_linesize,
yading@10:                   const uint8_t  *src, int src_linesize)
yading@10: {
yading@10:     int i, j;
yading@10:     EdgeDetectContext *edgedetect = ctx->priv;
yading@10: 
yading@10:     for (j = 1; j < h - 1; j++) {
yading@10:         dst += dst_linesize;
yading@10:         src += src_linesize;
yading@10:         for (i = 1; i < w - 1; i++) {
yading@10:             const int gx =
yading@10:                 -1*src[-src_linesize + i-1] + 1*src[-src_linesize + i+1]
yading@10:                 -2*src[                i-1] + 2*src[                i+1]
yading@10:                 -1*src[ src_linesize + i-1] + 1*src[ src_linesize + i+1];
yading@10:             const int gy =
yading@10:                 -1*src[-src_linesize + i-1] + 1*src[ src_linesize + i-1]
yading@10:                 -2*src[-src_linesize + i  ] + 2*src[ src_linesize + i  ]
yading@10:                 -1*src[-src_linesize + i+1] + 1*src[ src_linesize + i+1];
yading@10: 
yading@10:             dst[i] = FFABS(gx) + FFABS(gy);
yading@10:             edgedetect->directions[j*w + i] = get_rounded_direction(gx, gy);
yading@10:         }
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void non_maximum_suppression(AVFilterContext *ctx, int w, int h,
yading@10:                                           uint8_t  *dst, int dst_linesize,
yading@10:                                     const uint16_t *src, int src_linesize)
yading@10: {
yading@10:     int i, j;
yading@10:     EdgeDetectContext *edgedetect = ctx->priv;
yading@10: 
yading@10: #define COPY_MAXIMA(ay, ax, by, bx) do {                \
yading@10:     if (src[i] > src[(ay)*src_linesize + i+(ax)] &&     \
yading@10:         src[i] > src[(by)*src_linesize + i+(bx)])       \
yading@10:         dst[i] = av_clip_uint8(src[i]);                 \
yading@10: } while (0)
yading@10: 
yading@10:     for (j = 1; j < h - 1; j++) {
yading@10:         dst += dst_linesize;
yading@10:         src += src_linesize;
yading@10:         for (i = 1; i < w - 1; i++) {
yading@10:             switch (edgedetect->directions[j*w + i]) {
yading@10:             case DIRECTION_45UP:        COPY_MAXIMA( 1, -1, -1,  1); break;
yading@10:             case DIRECTION_45DOWN:      COPY_MAXIMA(-1, -1,  1,  1); break;
yading@10:             case DIRECTION_HORIZONTAL:  COPY_MAXIMA( 0, -1,  0,  1); break;
yading@10:             case DIRECTION_VERTICAL:    COPY_MAXIMA(-1,  0,  1,  0); break;
yading@10:             }
yading@10:         }
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void double_threshold(AVFilterContext *ctx, int w, int h,
yading@10:                                    uint8_t *dst, int dst_linesize,
yading@10:                              const uint8_t *src, int src_linesize)
yading@10: {
yading@10:     int i, j;
yading@10:     EdgeDetectContext *edgedetect = ctx->priv;
yading@10:     const int low  = edgedetect->low_u8;
yading@10:     const int high = edgedetect->high_u8;
yading@10: 
yading@10:     for (j = 0; j < h; j++) {
yading@10:         for (i = 0; i < w; i++) {
yading@10:             if (src[i] > high) {
yading@10:                 dst[i] = src[i];
yading@10:                 continue;
yading@10:             }
yading@10: 
yading@10:             if ((!i || i == w - 1 || !j || j == h - 1) &&
yading@10:                 src[i] > low &&
yading@10:                 (src[-src_linesize + i-1] > high ||
yading@10:                  src[-src_linesize + i  ] > high ||
yading@10:                  src[-src_linesize + i+1] > high ||
yading@10:                  src[                i-1] > high ||
yading@10:                  src[                i+1] > high ||
yading@10:                  src[ src_linesize + i-1] > high ||
yading@10:                  src[ src_linesize + i  ] > high ||
yading@10:                  src[ src_linesize + i+1] > high))
yading@10:                 dst[i] = src[i];
yading@10:             else
yading@10:                 dst[i] = 0;
yading@10:         }
yading@10:         dst += dst_linesize;
yading@10:         src += src_linesize;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static int filter_frame(AVFilterLink *inlink, AVFrame *in)
yading@10: {
yading@10:     AVFilterContext *ctx = inlink->dst;
yading@10:     EdgeDetectContext *edgedetect = ctx->priv;
yading@10:     AVFilterLink *outlink = inlink->dst->outputs[0];
yading@10:     uint8_t  *tmpbuf    = edgedetect->tmpbuf;
yading@10:     uint16_t *gradients = edgedetect->gradients;
yading@10:     int direct = 0;
yading@10:     AVFrame *out;
yading@10: 
yading@10:     if (av_frame_is_writable(in)) {
yading@10:         direct = 1;
yading@10:         out = in;
yading@10:     } else {
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: 
yading@10:     /* gaussian filter to reduce noise  */
yading@10:     gaussian_blur(ctx, inlink->w, inlink->h,
yading@10:                   tmpbuf,      inlink->w,
yading@10:                   in->data[0], in->linesize[0]);
yading@10: 
yading@10:     /* compute the 16-bits gradients and directions for the next step */
yading@10:     sobel(ctx, inlink->w, inlink->h,
yading@10:           gradients, inlink->w,
yading@10:           tmpbuf,    inlink->w);
yading@10: 
yading@10:     /* non_maximum_suppression() will actually keep & clip what's necessary and
yading@10:      * ignore the rest, so we need a clean output buffer */
yading@10:     memset(tmpbuf, 0, inlink->w * inlink->h);
yading@10:     non_maximum_suppression(ctx, inlink->w, inlink->h,
yading@10:                             tmpbuf,    inlink->w,
yading@10:                             gradients, inlink->w);
yading@10: 
yading@10:     /* keep high values, or low values surrounded by high values */
yading@10:     double_threshold(ctx, inlink->w, inlink->h,
yading@10:                      out->data[0], out->linesize[0],
yading@10:                      tmpbuf,       inlink->w);
yading@10: 
yading@10:     if (!direct)
yading@10:         av_frame_free(&in);
yading@10:     return ff_filter_frame(outlink, out);
yading@10: }
yading@10: 
yading@10: static av_cold void uninit(AVFilterContext *ctx)
yading@10: {
yading@10:     EdgeDetectContext *edgedetect = ctx->priv;
yading@10:     av_freep(&edgedetect->tmpbuf);
yading@10:     av_freep(&edgedetect->gradients);
yading@10:     av_freep(&edgedetect->directions);
yading@10: }
yading@10: 
yading@10: static const AVFilterPad edgedetect_inputs[] = {
yading@10:     {
yading@10:         .name         = "default",
yading@10:         .type         = AVMEDIA_TYPE_VIDEO,
yading@10:         .config_props = config_props,
yading@10:         .filter_frame = filter_frame,
yading@10:      },
yading@10:      { NULL }
yading@10: };
yading@10: 
yading@10: static const AVFilterPad edgedetect_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_edgedetect = {
yading@10:     .name          = "edgedetect",
yading@10:     .description   = NULL_IF_CONFIG_SMALL("Detect and draw edge."),
yading@10:     .priv_size     = sizeof(EdgeDetectContext),
yading@10:     .init          = init,
yading@10:     .uninit        = uninit,
yading@10:     .query_formats = query_formats,
yading@10:     .inputs        = edgedetect_inputs,
yading@10:     .outputs       = edgedetect_outputs,
yading@10:     .priv_class    = &edgedetect_class,
yading@10: };