[FFmpeg-devel] [PATCH 1/2] lafi/vf_edgedetect: Move some common functions into seperate file
Thilo Borgmann
thilo.borgmann at mail.de
Mon Apr 4 13:25:22 EEST 2022
Hi,
v3 updated to current HEAD.
-Thilo
-------------- next part --------------
From ba103f0a51557aae4f1c242ccc394da6ddce64fb Mon Sep 17 00:00:00 2001
From: Thilo Borgmann <thilo.borgmann at mail.de>
Date: Tue, 30 Nov 2021 00:16:52 +0100
Subject: [PATCH v3 1/2] lafi/vf_edgedetect: Move some common functions into
seperate file
---
libavfilter/Makefile | 2 +-
libavfilter/edge_common.c | 181 +++++++++++++++++++++++++++++++++++
libavfilter/edge_common.h | 107 +++++++++++++++++++++
libavfilter/vf_edgedetect.c | 183 ++----------------------------------
4 files changed, 296 insertions(+), 177 deletions(-)
create mode 100644 libavfilter/edge_common.c
create mode 100644 libavfilter/edge_common.h
diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index c4c946a988..006e59b2bd 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -265,7 +265,7 @@ OBJS-$(CONFIG_DRAWBOX_FILTER) += vf_drawbox.o
OBJS-$(CONFIG_DRAWGRAPH_FILTER) += f_drawgraph.o
OBJS-$(CONFIG_DRAWGRID_FILTER) += vf_drawbox.o
OBJS-$(CONFIG_DRAWTEXT_FILTER) += vf_drawtext.o
-OBJS-$(CONFIG_EDGEDETECT_FILTER) += vf_edgedetect.o
+OBJS-$(CONFIG_EDGEDETECT_FILTER) += vf_edgedetect.o edge_common.o
OBJS-$(CONFIG_ELBG_FILTER) += vf_elbg.o
OBJS-$(CONFIG_ENTROPY_FILTER) += vf_entropy.o
OBJS-$(CONFIG_EPX_FILTER) += vf_epx.o
diff --git a/libavfilter/edge_common.c b/libavfilter/edge_common.c
new file mode 100644
index 0000000000..d72e8521cd
--- /dev/null
+++ b/libavfilter/edge_common.c
@@ -0,0 +1,181 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "edge_common.h"
+
+// Internal helper for ff_sobel()
+static int get_rounded_direction(int gx, int gy)
+{
+ /* reference angles:
+ * tan( pi/8) = sqrt(2)-1
+ * tan(3pi/8) = sqrt(2)+1
+ * Gy/Gx is the tangent of the angle (theta), so Gy/Gx is compared against
+ * <ref-angle>, or more simply Gy against <ref-angle>*Gx
+ *
+ * Gx and Gy bounds = [-1020;1020], using 16-bit arithmetic:
+ * round((sqrt(2)-1) * (1<<16)) = 27146
+ * round((sqrt(2)+1) * (1<<16)) = 158218
+ */
+ if (gx) {
+ int tanpi8gx, tan3pi8gx;
+
+ if (gx < 0)
+ gx = -gx, gy = -gy;
+ gy *= (1 << 16);
+ tanpi8gx = 27146 * gx;
+ tan3pi8gx = 158218 * gx;
+ if (gy > -tan3pi8gx && gy < -tanpi8gx) return DIRECTION_45UP;
+ if (gy > -tanpi8gx && gy < tanpi8gx) return DIRECTION_HORIZONTAL;
+ if (gy > tanpi8gx && gy < tan3pi8gx) return DIRECTION_45DOWN;
+ }
+ return DIRECTION_VERTICAL;
+}
+
+// Simple sobel operator to get rounded gradients
+void ff_sobel(int w, int h,
+ uint16_t *dst, int dst_linesize,
+ int8_t *dir, int dir_linesize,
+ const uint8_t *src, int src_linesize)
+{
+ int i, j;
+
+ for (j = 1; j < h - 1; j++) {
+ dst += dst_linesize;
+ dir += dir_linesize;
+ src += src_linesize;
+ for (i = 1; i < w - 1; i++) {
+ const int gx =
+ -1*src[-src_linesize + i-1] + 1*src[-src_linesize + i+1]
+ -2*src[ i-1] + 2*src[ i+1]
+ -1*src[ src_linesize + i-1] + 1*src[ src_linesize + i+1];
+ const int gy =
+ -1*src[-src_linesize + i-1] + 1*src[ src_linesize + i-1]
+ -2*src[-src_linesize + i ] + 2*src[ src_linesize + i ]
+ -1*src[-src_linesize + i+1] + 1*src[ src_linesize + i+1];
+
+ dst[i] = FFABS(gx) + FFABS(gy);
+ dir[i] = get_rounded_direction(gx, gy);
+ }
+ }
+}
+
+// Filters rounded gradients to drop all non-maxima
+// Expects gradients generated by ff_sobel()
+// Expects zero's destination buffer
+void ff_non_maximum_suppression(int w, int h,
+ uint8_t *dst, int dst_linesize,
+ const int8_t *dir, int dir_linesize,
+ const uint16_t *src, int src_linesize)
+{
+ int i, j;
+
+#define COPY_MAXIMA(ay, ax, by, bx) do { \
+ if (src[i] > src[(ay)*src_linesize + i+(ax)] && \
+ src[i] > src[(by)*src_linesize + i+(bx)]) \
+ dst[i] = av_clip_uint8(src[i]); \
+} while (0)
+
+ for (j = 1; j < h - 1; j++) {
+ dst += dst_linesize;
+ dir += dir_linesize;
+ src += src_linesize;
+ for (i = 1; i < w - 1; i++) {
+ switch (dir[i]) {
+ case DIRECTION_45UP: COPY_MAXIMA( 1, -1, -1, 1); break;
+ case DIRECTION_45DOWN: COPY_MAXIMA(-1, -1, 1, 1); break;
+ case DIRECTION_HORIZONTAL: COPY_MAXIMA( 0, -1, 0, 1); break;
+ case DIRECTION_VERTICAL: COPY_MAXIMA(-1, 0, 1, 0); break;
+ }
+ }
+ }
+}
+
+// Filter to keep all pixels > high, and keep all pixels > low where all surrounding pixels > high
+void ff_double_threshold(int low, int high, int w, int h,
+ uint8_t *dst, int dst_linesize,
+ const uint8_t *src, int src_linesize)
+{
+ int i, j;
+
+ for (j = 0; j < h; j++) {
+ for (i = 0; i < w; i++) {
+ if (src[i] > high) {
+ dst[i] = src[i];
+ continue;
+ }
+
+ if (!(!i || i == w - 1 || !j || j == h - 1) &&
+ src[i] > low &&
+ (src[-src_linesize + i-1] > high ||
+ src[-src_linesize + i ] > high ||
+ src[-src_linesize + i+1] > high ||
+ src[ i-1] > high ||
+ src[ i+1] > high ||
+ src[ src_linesize + i-1] > high ||
+ src[ src_linesize + i ] > high ||
+ src[ src_linesize + i+1] > high))
+ dst[i] = src[i];
+ else
+ dst[i] = 0;
+ }
+ dst += dst_linesize;
+ src += src_linesize;
+ }
+}
+
+// Applies gaussian blur, using 5x5 kernels, sigma = 1.4
+void ff_gaussian_blur(int w, int h,
+ uint8_t *dst, int dst_linesize,
+ const uint8_t *src, int src_linesize)
+{
+ int i, j;
+
+ memcpy(dst, src, w); dst += dst_linesize; src += src_linesize;
+ memcpy(dst, src, w); dst += dst_linesize; src += src_linesize;
+ for (j = 2; j < h - 2; j++) {
+ dst[0] = src[0];
+ dst[1] = src[1];
+ for (i = 2; i < w - 2; i++) {
+ /* Gaussian mask of size 5x5 with sigma = 1.4 */
+ dst[i] = ((src[-2*src_linesize + i-2] + src[2*src_linesize + i-2]) * 2
+ + (src[-2*src_linesize + i-1] + src[2*src_linesize + i-1]) * 4
+ + (src[-2*src_linesize + i ] + src[2*src_linesize + i ]) * 5
+ + (src[-2*src_linesize + i+1] + src[2*src_linesize + i+1]) * 4
+ + (src[-2*src_linesize + i+2] + src[2*src_linesize + i+2]) * 2
+
+ + (src[ -src_linesize + i-2] + src[ src_linesize + i-2]) * 4
+ + (src[ -src_linesize + i-1] + src[ src_linesize + i-1]) * 9
+ + (src[ -src_linesize + i ] + src[ src_linesize + i ]) * 12
+ + (src[ -src_linesize + i+1] + src[ src_linesize + i+1]) * 9
+ + (src[ -src_linesize + i+2] + src[ src_linesize + i+2]) * 4
+
+ + src[i-2] * 5
+ + src[i-1] * 12
+ + src[i ] * 15
+ + src[i+1] * 12
+ + src[i+2] * 5) / 159;
+ }
+ dst[i ] = src[i ];
+ dst[i + 1] = src[i + 1];
+
+ dst += dst_linesize;
+ src += src_linesize;
+ }
+ memcpy(dst, src, w); dst += dst_linesize; src += src_linesize;
+ memcpy(dst, src, w);
+}
diff --git a/libavfilter/edge_common.h b/libavfilter/edge_common.h
new file mode 100644
index 0000000000..87c143f2b8
--- /dev/null
+++ b/libavfilter/edge_common.h
@@ -0,0 +1,107 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * common functions for edge detection
+ */
+
+#ifndef AVFILTER_EDGE_COMMON_H
+#define AVFILTER_EDGE_COMMON_H
+
+#include "avfilter.h"
+
+/**
+ * @brief Rounded directions used in av_image_sobel()
+ */
+enum AVRoundedDirection {
+ DIRECTION_45UP,
+ DIRECTION_45DOWN,
+ DIRECTION_HORIZONTAL,
+ DIRECTION_VERTICAL,
+};
+
+/**
+ * Simple sobel operator to get rounded gradients
+ *
+ * @param w the width of the image in pixels
+ * @param h the height of the image in pixels
+ * @param dst data pointers to magnitude image
+ * @param dst_linesize linesizes for the magnitude image
+ * @param dir data pointers to direction image
+ * @param dir_linesize linesizes for the direction image
+ * @param src data pointers to source image
+ * @param src_linesize linesizes for the source image
+ */
+void ff_sobel(int w, int h,
+ uint16_t *dst, int dst_linesize,
+ int8_t *dir, int dir_linesize,
+ const uint8_t *src, int src_linesize);
+
+/**
+ * Filters rounded gradients to drop all non-maxima pixels in the magnitude image
+ * Expects gradients generated by av_image_sobel()
+ * Expects zero's in the destination buffer dst
+ *
+ * @param w the width of the image in pixels
+ * @param h the height of the image in pixels
+ * @param dst data pointers to magnitude image
+ * @param dst_linesize linesizes for the magnitude image
+ * @param dir data pointers to direction image
+ * @param dir_linesize linesizes for the direction image
+ * @param src data pointers to source image
+ * @param src_linesize linesizes for the source image
+ */
+void ff_non_maximum_suppression(int w, int h,
+ uint8_t *dst, int dst_linesize,
+ const int8_t *dir, int dir_linesize,
+ const uint16_t *src, int src_linesize);
+
+/**
+ * Filters all pixels in src to keep all pixels > high,
+ * and keep all pixels > low where all surrounding pixels > high
+ *
+ * @param low the low threshold value
+ * @param high the hegh threshold value
+ * @param w the width of the image in pixels
+ * @param h the height of the image in pixels
+ * @param dst data pointers to destination image
+ * @param dst_linesize linesizes for the destination image
+ * @param src data pointers to source image
+ * @param src_linesize linesizes for the source image
+ */
+void ff_double_threshold(int low, int high, int w, int h,
+ uint8_t *dst, int dst_linesize,
+ const uint8_t *src, int src_linesize);
+
+/**
+ * Applies gaussian blur.
+ * 5x5 kernels, sigma = 1.4
+ *
+ * @param w the width of the image in pixels
+ * @param h the height of the image in pixels
+ * @param dst data pointers to destination image
+ * @param dst_linesize linesizes for the destination image
+ * @param src data pointers to source image
+ * @param src_linesize linesizes for the source image
+ */
+void ff_gaussian_blur(int w, int h,
+ uint8_t *dst, int dst_linesize,
+ const uint8_t *src, int src_linesize);
+
+#endif
diff --git a/libavfilter/vf_edgedetect.c b/libavfilter/vf_edgedetect.c
index 3eea34e325..90390ceb3e 100644
--- a/libavfilter/vf_edgedetect.c
+++ b/libavfilter/vf_edgedetect.c
@@ -32,6 +32,7 @@
#include "formats.h"
#include "internal.h"
#include "video.h"
+#include "edge_common.h"
#define PLANE_R 0x4
#define PLANE_G 0x1
@@ -139,176 +140,6 @@ static int config_props(AVFilterLink *inlink)
return 0;
}
-static void gaussian_blur(AVFilterContext *ctx, int w, int h,
- uint8_t *dst, int dst_linesize,
- const uint8_t *src, int src_linesize)
-{
- int i, j;
-
- memcpy(dst, src, w); dst += dst_linesize; src += src_linesize;
- if (h > 1) {
- memcpy(dst, src, w); dst += dst_linesize; src += src_linesize;
- }
- for (j = 2; j < h - 2; j++) {
- dst[0] = src[0];
- if (w > 1)
- dst[1] = src[1];
- for (i = 2; i < w - 2; i++) {
- /* Gaussian mask of size 5x5 with sigma = 1.4 */
- dst[i] = ((src[-2*src_linesize + i-2] + src[2*src_linesize + i-2]) * 2
- + (src[-2*src_linesize + i-1] + src[2*src_linesize + i-1]) * 4
- + (src[-2*src_linesize + i ] + src[2*src_linesize + i ]) * 5
- + (src[-2*src_linesize + i+1] + src[2*src_linesize + i+1]) * 4
- + (src[-2*src_linesize + i+2] + src[2*src_linesize + i+2]) * 2
-
- + (src[ -src_linesize + i-2] + src[ src_linesize + i-2]) * 4
- + (src[ -src_linesize + i-1] + src[ src_linesize + i-1]) * 9
- + (src[ -src_linesize + i ] + src[ src_linesize + i ]) * 12
- + (src[ -src_linesize + i+1] + src[ src_linesize + i+1]) * 9
- + (src[ -src_linesize + i+2] + src[ src_linesize + i+2]) * 4
-
- + src[i-2] * 5
- + src[i-1] * 12
- + src[i ] * 15
- + src[i+1] * 12
- + src[i+2] * 5) / 159;
- }
- if (w > 2)
- dst[i ] = src[i ];
- if (w > 3)
- dst[i + 1] = src[i + 1];
-
- dst += dst_linesize;
- src += src_linesize;
- }
- if (h > 2) {
- memcpy(dst, src, w); dst += dst_linesize; src += src_linesize;
- }
- if (h > 3)
- memcpy(dst, src, w);
-}
-
-enum {
- DIRECTION_45UP,
- DIRECTION_45DOWN,
- DIRECTION_HORIZONTAL,
- DIRECTION_VERTICAL,
-};
-
-static int get_rounded_direction(int gx, int gy)
-{
- /* reference angles:
- * tan( pi/8) = sqrt(2)-1
- * tan(3pi/8) = sqrt(2)+1
- * Gy/Gx is the tangent of the angle (theta), so Gy/Gx is compared against
- * <ref-angle>, or more simply Gy against <ref-angle>*Gx
- *
- * Gx and Gy bounds = [-1020;1020], using 16-bit arithmetic:
- * round((sqrt(2)-1) * (1<<16)) = 27146
- * round((sqrt(2)+1) * (1<<16)) = 158218
- */
- if (gx) {
- int tanpi8gx, tan3pi8gx;
-
- if (gx < 0)
- gx = -gx, gy = -gy;
- gy *= (1 << 16);
- tanpi8gx = 27146 * gx;
- tan3pi8gx = 158218 * gx;
- if (gy > -tan3pi8gx && gy < -tanpi8gx) return DIRECTION_45UP;
- if (gy > -tanpi8gx && gy < tanpi8gx) return DIRECTION_HORIZONTAL;
- if (gy > tanpi8gx && gy < tan3pi8gx) return DIRECTION_45DOWN;
- }
- return DIRECTION_VERTICAL;
-}
-
-static void sobel(int w, int h,
- uint16_t *dst, int dst_linesize,
- int8_t *dir, int dir_linesize,
- const uint8_t *src, int src_linesize)
-{
- int i, j;
-
- for (j = 1; j < h - 1; j++) {
- dst += dst_linesize;
- dir += dir_linesize;
- src += src_linesize;
- for (i = 1; i < w - 1; i++) {
- const int gx =
- -1*src[-src_linesize + i-1] + 1*src[-src_linesize + i+1]
- -2*src[ i-1] + 2*src[ i+1]
- -1*src[ src_linesize + i-1] + 1*src[ src_linesize + i+1];
- const int gy =
- -1*src[-src_linesize + i-1] + 1*src[ src_linesize + i-1]
- -2*src[-src_linesize + i ] + 2*src[ src_linesize + i ]
- -1*src[-src_linesize + i+1] + 1*src[ src_linesize + i+1];
-
- dst[i] = FFABS(gx) + FFABS(gy);
- dir[i] = get_rounded_direction(gx, gy);
- }
- }
-}
-
-static void non_maximum_suppression(int w, int h,
- uint8_t *dst, int dst_linesize,
- const int8_t *dir, int dir_linesize,
- const uint16_t *src, int src_linesize)
-{
- int i, j;
-
-#define COPY_MAXIMA(ay, ax, by, bx) do { \
- if (src[i] > src[(ay)*src_linesize + i+(ax)] && \
- src[i] > src[(by)*src_linesize + i+(bx)]) \
- dst[i] = av_clip_uint8(src[i]); \
-} while (0)
-
- for (j = 1; j < h - 1; j++) {
- dst += dst_linesize;
- dir += dir_linesize;
- src += src_linesize;
- for (i = 1; i < w - 1; i++) {
- switch (dir[i]) {
- case DIRECTION_45UP: COPY_MAXIMA( 1, -1, -1, 1); break;
- case DIRECTION_45DOWN: COPY_MAXIMA(-1, -1, 1, 1); break;
- case DIRECTION_HORIZONTAL: COPY_MAXIMA( 0, -1, 0, 1); break;
- case DIRECTION_VERTICAL: COPY_MAXIMA(-1, 0, 1, 0); break;
- }
- }
- }
-}
-
-static void double_threshold(int low, int high, int w, int h,
- uint8_t *dst, int dst_linesize,
- const uint8_t *src, int src_linesize)
-{
- int i, j;
-
- for (j = 0; j < h; j++) {
- for (i = 0; i < w; i++) {
- if (src[i] > high) {
- dst[i] = src[i];
- continue;
- }
-
- if (!(!i || i == w - 1 || !j || j == h - 1) &&
- src[i] > low &&
- (src[-src_linesize + i-1] > high ||
- src[-src_linesize + i ] > high ||
- src[-src_linesize + i+1] > high ||
- src[ i-1] > high ||
- src[ i+1] > high ||
- src[ src_linesize + i-1] > high ||
- src[ src_linesize + i ] > high ||
- src[ src_linesize + i+1] > high))
- dst[i] = src[i];
- else
- dst[i] = 0;
- }
- dst += dst_linesize;
- src += src_linesize;
- }
-}
-
static void color_mix(int w, int h,
uint8_t *dst, int dst_linesize,
const uint8_t *src, int src_linesize)
@@ -360,12 +191,12 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *in)
}
/* gaussian filter to reduce noise */
- gaussian_blur(ctx, width, height,
- tmpbuf, width,
- in->data[p], in->linesize[p]);
+ ff_gaussian_blur(width, height,
+ tmpbuf, width,
+ in->data[p], in->linesize[p]);
/* compute the 16-bits gradients and directions for the next step */
- sobel(width, height,
+ ff_sobel(width, height,
gradients, width,
directions,width,
tmpbuf, width);
@@ -373,13 +204,13 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *in)
/* non_maximum_suppression() will actually keep & clip what's necessary and
* ignore the rest, so we need a clean output buffer */
memset(tmpbuf, 0, width * height);
- non_maximum_suppression(width, height,
+ ff_non_maximum_suppression(width, height,
tmpbuf, width,
directions,width,
gradients, width);
/* keep high values, or low values surrounded by high values */
- double_threshold(edgedetect->low_u8, edgedetect->high_u8,
+ ff_double_threshold(edgedetect->low_u8, edgedetect->high_u8,
width, height,
out->data[p], out->linesize[p],
tmpbuf, width);
--
2.20.1 (Apple Git-117)
More information about the ffmpeg-devel
mailing list