The code is based on the remove-logo filter in MPlayer/libmpcodecs, by
Robert Edele, relicensed to LGPL with consent of the author.
Address trac issue #249.
... | ... |
@@ -1688,6 +1688,7 @@ mptestsrc_filter_deps="gpl" |
1688 | 1688 |
negate_filter_deps="lut_filter" |
1689 | 1689 |
ocv_filter_deps="libopencv" |
1690 | 1690 |
pan_filter_deps="swresample" |
1691 |
+removelogo_filter_deps="avcodec avformat swscale" |
|
1691 | 1692 |
scale_filter_deps="swscale" |
1692 | 1693 |
tinterlace_filter_deps="gpl" |
1693 | 1694 |
yadif_filter_deps="gpl" |
... | ... |
@@ -2237,6 +2237,32 @@ format=monow, pixdesctest |
2237 | 2237 |
|
2238 | 2238 |
can be used to test the monowhite pixel format descriptor definition. |
2239 | 2239 |
|
2240 |
+@section removelogo |
|
2241 |
+ |
|
2242 |
+Suppress a TV station logo, using an image file to determine which |
|
2243 |
+pixels comprise the logo. It works by filling in the pixels that |
|
2244 |
+comprise the logo with neighboring pixels. |
|
2245 |
+ |
|
2246 |
+This filter requires one argument which specifies the filter bitmap |
|
2247 |
+file, which can be any image format supported by libavformat. The |
|
2248 |
+width and height of the image file must match those of the video |
|
2249 |
+stream being processed. |
|
2250 |
+ |
|
2251 |
+Pixels in the provided bitmap image with a value of zero are not |
|
2252 |
+considered part of the logo, non-zero pixels are considered part of |
|
2253 |
+the logo. If you use white (255) for the logo and black (0) for the |
|
2254 |
+rest, you will be safe. For making the filter bitmap, it is |
|
2255 |
+recommended to take a screen capture of a black frame with the logo |
|
2256 |
+visible, and then using a threshold filter followed by the erode |
|
2257 |
+filter once or twice. |
|
2258 |
+ |
|
2259 |
+If needed, little splotches can be fixed manually. Remember that if |
|
2260 |
+logo pixels are not covered, the filter quality will be much |
|
2261 |
+reduced. Marking too many pixels as part of the logo does not hurt as |
|
2262 |
+much, but it will increase the amount of blurring needed to cover over |
|
2263 |
+the image and will destroy more information than necessary, and extra |
|
2264 |
+pixels will slow things down on a large logo. |
|
2265 |
+ |
|
2240 | 2266 |
@section scale |
2241 | 2267 |
|
2242 | 2268 |
Scale the input video to @var{width}:@var{height}[:@var{interl}=@{1|-1@}] and/or convert the image format. |
... | ... |
@@ -8,6 +8,7 @@ FFLIBS-$(CONFIG_AMOVIE_FILTER) += avformat avcodec |
8 | 8 |
FFLIBS-$(CONFIG_ARESAMPLE_FILTER) += swresample |
9 | 9 |
FFLIBS-$(CONFIG_MOVIE_FILTER) += avformat avcodec |
10 | 10 |
FFLIBS-$(CONFIG_PAN_FILTER) += swresample |
11 |
+FFLIBS-$(CONFIG_REMOVELOGO_FILTER) += swscale avformat avcodec |
|
11 | 12 |
FFLIBS-$(CONFIG_SCALE_FILTER) += swscale |
12 | 13 |
FFLIBS-$(CONFIG_MP_FILTER) += avcodec postproc |
13 | 14 |
|
... | ... |
@@ -25,6 +26,8 @@ OBJS = allfilters.o \ |
25 | 25 |
vsrc_buffer.o |
26 | 26 |
|
27 | 27 |
OBJS-$(CONFIG_AVCODEC) += avcodec.o |
28 |
+OBJS-$(CONFIG_AVFORMAT) += lavfutils.o |
|
29 |
+OBJS-$(CONFIG_SWSCALE) += lswsutils.o |
|
28 | 30 |
|
29 | 31 |
OBJS-$(CONFIG_ACONVERT_FILTER) += af_aconvert.o |
30 | 32 |
OBJS-$(CONFIG_AFORMAT_FILTER) += af_aformat.o |
... | ... |
@@ -80,6 +83,7 @@ OBJS-$(CONFIG_OCV_FILTER) += vf_libopencv.o |
80 | 80 |
OBJS-$(CONFIG_OVERLAY_FILTER) += vf_overlay.o |
81 | 81 |
OBJS-$(CONFIG_PAD_FILTER) += vf_pad.o |
82 | 82 |
OBJS-$(CONFIG_PIXDESCTEST_FILTER) += vf_pixdesctest.o |
83 |
+OBJS-$(CONFIG_REMOVELOGO_FILTER) += bbox.o lswsutils.o lavfutils.o vf_removelogo.o |
|
83 | 84 |
OBJS-$(CONFIG_SCALE_FILTER) += vf_scale.o |
84 | 85 |
OBJS-$(CONFIG_SELECT_FILTER) += vf_select.o |
85 | 86 |
OBJS-$(CONFIG_SETDAR_FILTER) += vf_aspect.o |
... | ... |
@@ -88,6 +88,7 @@ void avfilter_register_all(void) |
88 | 88 |
REGISTER_FILTER (OVERLAY, overlay, vf); |
89 | 89 |
REGISTER_FILTER (PAD, pad, vf); |
90 | 90 |
REGISTER_FILTER (PIXDESCTEST, pixdesctest, vf); |
91 |
+ REGISTER_FILTER (REMOVELOGO, removelogo, vf); |
|
91 | 92 |
REGISTER_FILTER (SCALE, scale, vf); |
92 | 93 |
REGISTER_FILTER (SELECT, select, vf); |
93 | 94 |
REGISTER_FILTER (SETDAR, setdar, vf); |
94 | 95 |
new file mode 100644 |
... | ... |
@@ -0,0 +1,95 @@ |
0 |
+/* |
|
1 |
+ * This file is part of FFmpeg. |
|
2 |
+ * |
|
3 |
+ * FFmpeg is free software; you can redistribute it and/or |
|
4 |
+ * modify it under the terms of the GNU Lesser General Public |
|
5 |
+ * License as published by the Free Software Foundation; either |
|
6 |
+ * version 2.1 of the License, or (at your option) any later version. |
|
7 |
+ * |
|
8 |
+ * FFmpeg is distributed in the hope that it will be useful, |
|
9 |
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
10 |
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|
11 |
+ * Lesser General Public License for more details. |
|
12 |
+ * |
|
13 |
+ * You should have received a copy of the GNU Lesser General Public |
|
14 |
+ * License along with FFmpeg; if not, write to the Free Software |
|
15 |
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
|
16 |
+ */ |
|
17 |
+ |
|
18 |
+#include "libavutil/imgutils.h" |
|
19 |
+#include "lavfutils.h" |
|
20 |
+ |
|
21 |
+int ff_load_image(uint8_t *data[4], int linesize[4], |
|
22 |
+ int *w, int *h, enum PixelFormat *pix_fmt, |
|
23 |
+ const char *filename, void *log_ctx) |
|
24 |
+{ |
|
25 |
+ AVInputFormat *iformat = NULL; |
|
26 |
+ AVFormatContext *format_ctx; |
|
27 |
+ AVCodec *codec; |
|
28 |
+ AVCodecContext *codec_ctx; |
|
29 |
+ AVFrame *frame; |
|
30 |
+ int frame_decoded, ret = 0; |
|
31 |
+ AVPacket pkt; |
|
32 |
+ |
|
33 |
+ av_register_all(); |
|
34 |
+ |
|
35 |
+ iformat = av_find_input_format("image2"); |
|
36 |
+ if ((ret = avformat_open_input(&format_ctx, filename, iformat, NULL)) < 0) { |
|
37 |
+ av_log(log_ctx, AV_LOG_ERROR, |
|
38 |
+ "Failed to open input file '%s'\n", filename); |
|
39 |
+ return ret; |
|
40 |
+ } |
|
41 |
+ |
|
42 |
+ codec_ctx = format_ctx->streams[0]->codec; |
|
43 |
+ codec = avcodec_find_decoder(codec_ctx->codec_id); |
|
44 |
+ if (!codec) { |
|
45 |
+ av_log(log_ctx, AV_LOG_ERROR, "Failed to find codec\n"); |
|
46 |
+ ret = AVERROR(EINVAL); |
|
47 |
+ goto end; |
|
48 |
+ } |
|
49 |
+ |
|
50 |
+ if ((ret = avcodec_open2(codec_ctx, codec, NULL)) < 0) { |
|
51 |
+ av_log(log_ctx, AV_LOG_ERROR, "Failed to open codec\n"); |
|
52 |
+ goto end; |
|
53 |
+ } |
|
54 |
+ |
|
55 |
+ if (!(frame = avcodec_alloc_frame()) ) { |
|
56 |
+ av_log(log_ctx, AV_LOG_ERROR, "Failed to alloc frame\n"); |
|
57 |
+ ret = AVERROR(ENOMEM); |
|
58 |
+ goto end; |
|
59 |
+ } |
|
60 |
+ |
|
61 |
+ ret = av_read_frame(format_ctx, &pkt); |
|
62 |
+ if (ret < 0) { |
|
63 |
+ av_log(log_ctx, AV_LOG_ERROR, "Failed to read frame from file\n"); |
|
64 |
+ goto end; |
|
65 |
+ } |
|
66 |
+ |
|
67 |
+ ret = avcodec_decode_video2(codec_ctx, frame, &frame_decoded, &pkt); |
|
68 |
+ if (ret < 0 || !frame_decoded) { |
|
69 |
+ av_log(log_ctx, AV_LOG_ERROR, "Failed to decode image from file\n"); |
|
70 |
+ goto end; |
|
71 |
+ } |
|
72 |
+ ret = 0; |
|
73 |
+ |
|
74 |
+ *w = frame->width; |
|
75 |
+ *h = frame->height; |
|
76 |
+ *pix_fmt = frame->format; |
|
77 |
+ |
|
78 |
+ if ((ret = av_image_alloc(data, linesize, *w, *h, *pix_fmt, 16)) < 0) |
|
79 |
+ goto end; |
|
80 |
+ ret = 0; |
|
81 |
+ |
|
82 |
+ av_image_copy(data, linesize, frame->data, frame->linesize, *pix_fmt, *w, *h); |
|
83 |
+ |
|
84 |
+end: |
|
85 |
+ if (codec_ctx) |
|
86 |
+ avcodec_close(codec_ctx); |
|
87 |
+ if (format_ctx) |
|
88 |
+ avformat_close_input(&format_ctx); |
|
89 |
+ av_freep(&frame); |
|
90 |
+ |
|
91 |
+ if (ret < 0) |
|
92 |
+ av_log(log_ctx, AV_LOG_ERROR, "Error loading image file '%s'\n", filename); |
|
93 |
+ return ret; |
|
94 |
+} |
0 | 95 |
new file mode 100644 |
... | ... |
@@ -0,0 +1,43 @@ |
0 |
+/* |
|
1 |
+ * This file is part of FFmpeg. |
|
2 |
+ * |
|
3 |
+ * FFmpeg is free software; you can redistribute it and/or |
|
4 |
+ * modify it under the terms of the GNU Lesser General Public |
|
5 |
+ * License as published by the Free Software Foundation; either |
|
6 |
+ * version 2.1 of the License, or (at your option) any later version. |
|
7 |
+ * |
|
8 |
+ * FFmpeg is distributed in the hope that it will be useful, |
|
9 |
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
10 |
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|
11 |
+ * Lesser General Public License for more details. |
|
12 |
+ * |
|
13 |
+ * You should have received a copy of the GNU Lesser General Public |
|
14 |
+ * License along with FFmpeg; if not, write to the Free Software |
|
15 |
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
|
16 |
+ */ |
|
17 |
+ |
|
18 |
+/** |
|
19 |
+ * @file |
|
20 |
+ * Miscellaneous utilities which make use of the libavformat library |
|
21 |
+ */ |
|
22 |
+ |
|
23 |
+#ifndef AVFILTER_LAVFUTILS_H |
|
24 |
+#define AVFILTER_LAVFUTILS_H |
|
25 |
+ |
|
26 |
+#include "libavformat/avformat.h" |
|
27 |
+ |
|
28 |
+/** |
|
29 |
+ * Load image from filename and put the resulting image in data. |
|
30 |
+ * |
|
31 |
+ * @param w pointer to the width of the loaded image |
|
32 |
+ * @param h pointer to the height of the loaded image |
|
33 |
+ * @param pix_fmt pointer to the pixel format of the loaded image |
|
34 |
+ * @param filename the name of the image file to load |
|
35 |
+ * @param log_ctx log context |
|
36 |
+ * @return 0 in case of success, a negative error code otherwise. |
|
37 |
+ */ |
|
38 |
+int ff_load_image(uint8_t *data[4], int linesize[4], |
|
39 |
+ int *w, int *h, enum PixelFormat *pix_fmt, |
|
40 |
+ const char *filename, void *log_ctx); |
|
41 |
+ |
|
42 |
+#endif /* AVFILTER_LAVFUTILS_H */ |
0 | 43 |
new file mode 100644 |
... | ... |
@@ -0,0 +1,50 @@ |
0 |
+/* |
|
1 |
+ * This file is part of FFmpeg. |
|
2 |
+ * |
|
3 |
+ * FFmpeg is free software; you can redistribute it and/or |
|
4 |
+ * modify it under the terms of the GNU Lesser General Public |
|
5 |
+ * License as published by the Free Software Foundation; either |
|
6 |
+ * version 2.1 of the License, or (at your option) any later version. |
|
7 |
+ * |
|
8 |
+ * FFmpeg is distributed in the hope that it will be useful, |
|
9 |
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
10 |
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|
11 |
+ * Lesser General Public License for more details. |
|
12 |
+ * |
|
13 |
+ * You should have received a copy of the GNU Lesser General Public |
|
14 |
+ * License along with FFmpeg; if not, write to the Free Software |
|
15 |
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
|
16 |
+ */ |
|
17 |
+ |
|
18 |
+#include "libavutil/imgutils.h" |
|
19 |
+#include "lswsutils.h" |
|
20 |
+ |
|
21 |
+int ff_scale_image(uint8_t *dst_data[4], int dst_linesize[4], |
|
22 |
+ int dst_w, int dst_h, enum PixelFormat dst_pix_fmt, |
|
23 |
+ uint8_t * const src_data[4], int src_linesize[4], |
|
24 |
+ int src_w, int src_h, enum PixelFormat src_pix_fmt, |
|
25 |
+ void *log_ctx) |
|
26 |
+{ |
|
27 |
+ int ret; |
|
28 |
+ struct SwsContext *sws_ctx = sws_getContext(src_w, src_h, src_pix_fmt, |
|
29 |
+ dst_w, dst_h, dst_pix_fmt, |
|
30 |
+ SWS_BILINEAR, NULL, NULL, NULL); |
|
31 |
+ if (!sws_ctx) { |
|
32 |
+ av_log(log_ctx, AV_LOG_ERROR, |
|
33 |
+ "Impossible to create scale context for the conversion " |
|
34 |
+ "fmt:%s s:%dx%d -> fmt:%s s:%dx%d\n", |
|
35 |
+ av_get_pix_fmt_name(src_pix_fmt), src_w, src_h, |
|
36 |
+ av_get_pix_fmt_name(dst_pix_fmt), dst_w, dst_h); |
|
37 |
+ ret = AVERROR(EINVAL); |
|
38 |
+ goto end; |
|
39 |
+ } |
|
40 |
+ |
|
41 |
+ if ((ret = av_image_alloc(dst_data, dst_linesize, dst_w, dst_h, dst_pix_fmt, 16)) < 0) |
|
42 |
+ goto end; |
|
43 |
+ ret = 0; |
|
44 |
+ sws_scale(sws_ctx, src_data, src_linesize, 0, src_h, dst_data, dst_linesize); |
|
45 |
+ |
|
46 |
+end: |
|
47 |
+ sws_freeContext(sws_ctx); |
|
48 |
+ return ret; |
|
49 |
+} |
0 | 50 |
new file mode 100644 |
... | ... |
@@ -0,0 +1,38 @@ |
0 |
+/* |
|
1 |
+ * This file is part of FFmpeg. |
|
2 |
+ * |
|
3 |
+ * FFmpeg is free software; you can redistribute it and/or |
|
4 |
+ * modify it under the terms of the GNU Lesser General Public |
|
5 |
+ * License as published by the Free Software Foundation; either |
|
6 |
+ * version 2.1 of the License, or (at your option) any later version. |
|
7 |
+ * |
|
8 |
+ * FFmpeg is distributed in the hope that it will be useful, |
|
9 |
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
10 |
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|
11 |
+ * Lesser General Public License for more details. |
|
12 |
+ * |
|
13 |
+ * You should have received a copy of the GNU Lesser General Public |
|
14 |
+ * License along with FFmpeg; if not, write to the Free Software |
|
15 |
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
|
16 |
+ */ |
|
17 |
+ |
|
18 |
+/** |
|
19 |
+ * @file |
|
20 |
+ * Miscellaneous utilities which make use of the libswscale library |
|
21 |
+ */ |
|
22 |
+ |
|
23 |
+#ifndef AVFILTER_LSWSUTILS_H |
|
24 |
+#define AVFILTER_LSWSUTILS_H |
|
25 |
+ |
|
26 |
+#include "libswscale/swscale.h" |
|
27 |
+ |
|
28 |
+/** |
|
29 |
+ * Scale image using libswscale. |
|
30 |
+ */ |
|
31 |
+int ff_scale_image(uint8_t *dst_data[4], int dst_linesize[4], |
|
32 |
+ int dst_w, int dst_h, enum PixelFormat dst_pix_fmt, |
|
33 |
+ uint8_t *const src_data[4], int src_linesize[4], |
|
34 |
+ int src_w, int src_h, enum PixelFormat src_pix_fmt, |
|
35 |
+ void *log_ctx); |
|
36 |
+ |
|
37 |
+#endif /* AVFILTER_LSWSUTILS_H */ |
... | ... |
@@ -29,8 +29,8 @@ |
29 | 29 |
#include "libavutil/avutil.h" |
30 | 30 |
|
31 | 31 |
#define LIBAVFILTER_VERSION_MAJOR 2 |
32 |
-#define LIBAVFILTER_VERSION_MINOR 66 |
|
33 |
-#define LIBAVFILTER_VERSION_MICRO 101 |
|
32 |
+#define LIBAVFILTER_VERSION_MINOR 67 |
|
33 |
+#define LIBAVFILTER_VERSION_MICRO 100 |
|
34 | 34 |
|
35 | 35 |
#define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \ |
36 | 36 |
LIBAVFILTER_VERSION_MINOR, \ |
37 | 37 |
new file mode 100644 |
... | ... |
@@ -0,0 +1,570 @@ |
0 |
+/* |
|
1 |
+ * Copyright (c) 2005 Robert Edele <yartrebo@earthlink.net> |
|
2 |
+ * Copyright (c) 2012 Stefano Sabatini |
|
3 |
+ * |
|
4 |
+ * This file is part of FFmpeg. |
|
5 |
+ * |
|
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. |
|
10 |
+ * |
|
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. |
|
15 |
+ * |
|
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 |
|
19 |
+ */ |
|
20 |
+ |
|
21 |
+/** |
|
22 |
+ * @file advanced blur-based logo removing filter |
|
23 |
+ * |
|
24 |
+ * This filter loads an image mask file showing where a logo is and |
|
25 |
+ * uses a blur transform to remove the logo. |
|
26 |
+ * |
|
27 |
+ * Based on the libmpcodecs remove-logo filter by Robert Edele. |
|
28 |
+ */ |
|
29 |
+ |
|
30 |
+/** |
|
31 |
+ * This code implements a filter to remove annoying TV logos and other annoying |
|
32 |
+ * images placed onto a video stream. It works by filling in the pixels that |
|
33 |
+ * comprise the logo with neighboring pixels. The transform is very loosely |
|
34 |
+ * based on a gaussian blur, but it is different enough to merit its own |
|
35 |
+ * paragraph later on. It is a major improvement on the old delogo filter as it |
|
36 |
+ * both uses a better blurring algorithm and uses a bitmap to use an arbitrary |
|
37 |
+ * and generally much tighter fitting shape than a rectangle. |
|
38 |
+ * |
|
39 |
+ * The logo removal algorithm has two key points. The first is that it |
|
40 |
+ * distinguishes between pixels in the logo and those not in the logo by using |
|
41 |
+ * the passed-in bitmap. Pixels not in the logo are copied over directly without |
|
42 |
+ * being modified and they also serve as source pixels for the logo |
|
43 |
+ * fill-in. Pixels inside the logo have the mask applied. |
|
44 |
+ * |
|
45 |
+ * At init-time the bitmap is reprocessed internally, and the distance to the |
|
46 |
+ * nearest edge of the logo (Manhattan distance), along with a little extra to |
|
47 |
+ * remove rough edges, is stored in each pixel. This is done using an in-place |
|
48 |
+ * erosion algorithm, and incrementing each pixel that survives any given |
|
49 |
+ * erosion. Once every pixel is eroded, the maximum value is recorded, and a |
|
50 |
+ * set of masks from size 0 to this size are generaged. The masks are circular |
|
51 |
+ * binary masks, where each pixel within a radius N (where N is the size of the |
|
52 |
+ * mask) is a 1, and all other pixels are a 0. Although a gaussian mask would be |
|
53 |
+ * more mathematically accurate, a binary mask works better in practice because |
|
54 |
+ * we generally do not use the central pixels in the mask (because they are in |
|
55 |
+ * the logo region), and thus a gaussian mask will cause too little blur and |
|
56 |
+ * thus a very unstable image. |
|
57 |
+ * |
|
58 |
+ * The mask is applied in a special way. Namely, only pixels in the mask that |
|
59 |
+ * line up to pixels outside the logo are used. The dynamic mask size means that |
|
60 |
+ * the mask is just big enough so that the edges touch pixels outside the logo, |
|
61 |
+ * so the blurring is kept to a minimum and at least the first boundary |
|
62 |
+ * condition is met (that the image function itself is continuous), even if the |
|
63 |
+ * second boundary condition (that the derivative of the image function is |
|
64 |
+ * continuous) is not met. A masking algorithm that does preserve the second |
|
65 |
+ * boundary coundition (perhaps something based on a highly-modified bi-cubic |
|
66 |
+ * algorithm) should offer even better results on paper, but the noise in a |
|
67 |
+ * typical TV signal should make anything based on derivatives hopelessly noisy. |
|
68 |
+ */ |
|
69 |
+ |
|
70 |
+#include "libavutil/imgutils.h" |
|
71 |
+#include "avfilter.h" |
|
72 |
+#include "bbox.h" |
|
73 |
+#include "lavfutils.h" |
|
74 |
+#include "lswsutils.h" |
|
75 |
+ |
|
76 |
+typedef struct { |
|
77 |
+ /* Stores our collection of masks. The first is for an array of |
|
78 |
+ the second for the y axis, and the third for the x axis. */ |
|
79 |
+ int ***mask; |
|
80 |
+ int max_mask_size; |
|
81 |
+ int mask_w, mask_h; |
|
82 |
+ |
|
83 |
+ uint8_t *full_mask_data; |
|
84 |
+ FFBoundingBox full_mask_bbox; |
|
85 |
+ uint8_t *half_mask_data; |
|
86 |
+ FFBoundingBox half_mask_bbox; |
|
87 |
+} RemovelogoContext; |
|
88 |
+ |
|
89 |
+/** |
|
90 |
+ * Choose a slightly larger mask size to improve performance. |
|
91 |
+ * |
|
92 |
+ * This function maps the absolute minimum mask size needed to the |
|
93 |
+ * mask size we'll actually use. f(x) = x (the smallest that will |
|
94 |
+ * work) will produce the sharpest results, but will be quite |
|
95 |
+ * jittery. f(x) = 1.25x (what I'm using) is a good tradeoff in my |
|
96 |
+ * opinion. This will calculate only at init-time, so you can put a |
|
97 |
+ * long expression here without effecting performance. |
|
98 |
+ */ |
|
99 |
+#define apply_mask_fudge_factor(x) (((x) >> 2) + x) |
|
100 |
+ |
|
101 |
+/** |
|
102 |
+ * Pre-process an image to give distance information. |
|
103 |
+ * |
|
104 |
+ * This function takes a bitmap image and converts it in place into a |
|
105 |
+ * distance image. A distance image is zero for pixels outside of the |
|
106 |
+ * logo and is the Manhattan distance (|dx| + |dy|) from the logo edge |
|
107 |
+ * for pixels inside of the logo. This will overestimate the distance, |
|
108 |
+ * but that is safe, and is far easier to implement than a proper |
|
109 |
+ * pythagorean distance since I'm using a modified erosion algorithm |
|
110 |
+ * to compute the distances. |
|
111 |
+ * |
|
112 |
+ * @param mask image which will be converted from a greyscale image |
|
113 |
+ * into a distance image. |
|
114 |
+ */ |
|
115 |
+static void convert_mask_to_strength_mask(uint8_t *data, int linesize, |
|
116 |
+ int w, int h, int min_val, |
|
117 |
+ int *max_mask_size) |
|
118 |
+{ |
|
119 |
+ int x, y; |
|
120 |
+ |
|
121 |
+ /* How many times we've gone through the loop. Used in the |
|
122 |
+ in-place erosion algorithm and to get us max_mask_size later on. */ |
|
123 |
+ int current_pass = 0; |
|
124 |
+ |
|
125 |
+ /* set all non-zero values to 1 */ |
|
126 |
+ for (y = 0; y < h; y++) |
|
127 |
+ for (x = 0; x < w; x++) |
|
128 |
+ data[y*linesize + x] = data[y*linesize + x] > min_val; |
|
129 |
+ |
|
130 |
+ /* For each pass, if a pixel is itself the same value as the |
|
131 |
+ current pass, and its four neighbors are too, then it is |
|
132 |
+ incremented. If no pixels are incremented by the end of the |
|
133 |
+ pass, then we go again. Edge pixels are counted as always |
|
134 |
+ excluded (this should be true anyway for any sane mask, but if |
|
135 |
+ it isn't this will ensure that we eventually exit). */ |
|
136 |
+ while (1) { |
|
137 |
+ /* If this doesn't get set by the end of this pass, then we're done. */ |
|
138 |
+ int has_anything_changed = 0; |
|
139 |
+ uint8_t *current_pixel0 = data, *current_pixel; |
|
140 |
+ current_pass++; |
|
141 |
+ |
|
142 |
+ for (y = 1; y < h-1; y++) { |
|
143 |
+ current_pixel = current_pixel0; |
|
144 |
+ for (x = 1; x < w-1; x++) { |
|
145 |
+ /* Apply the in-place erosion transform. It is based |
|
146 |
+ on the following two premises: |
|
147 |
+ 1 - Any pixel that fails 1 erosion will fail all |
|
148 |
+ future erosions. |
|
149 |
+ |
|
150 |
+ 2 - Only pixels having survived all erosions up to |
|
151 |
+ the present will be >= to current_pass. |
|
152 |
+ It doesn't matter if it survived the current pass, |
|
153 |
+ failed it, or hasn't been tested yet. By using >= |
|
154 |
+ instead of ==, we allow the algorithm to work in |
|
155 |
+ place. */ |
|
156 |
+ if ( *current_pixel >= current_pass && |
|
157 |
+ *(current_pixel + 1) >= current_pass && |
|
158 |
+ *(current_pixel - 1) >= current_pass && |
|
159 |
+ *(current_pixel + w) >= current_pass && |
|
160 |
+ *(current_pixel - w) >= current_pass) { |
|
161 |
+ /* Increment the value since it still has not been |
|
162 |
+ * eroded, as evidenced by the if statement that |
|
163 |
+ * just evaluated to true. */ |
|
164 |
+ (*current_pixel)++; |
|
165 |
+ has_anything_changed = 1; |
|
166 |
+ } |
|
167 |
+ current_pixel++; |
|
168 |
+ } |
|
169 |
+ current_pixel0 += linesize; |
|
170 |
+ } |
|
171 |
+ if (!has_anything_changed) |
|
172 |
+ break; |
|
173 |
+ } |
|
174 |
+ |
|
175 |
+ /* Apply the fudge factor, which will increase the size of the |
|
176 |
+ * mask a little to reduce jitter at the cost of more blur. */ |
|
177 |
+ for (y = 1; y < h - 1; y++) |
|
178 |
+ for (x = 1; x < w - 1; x++) |
|
179 |
+ data[(y * linesize) + x] = apply_mask_fudge_factor(data[(y * linesize) + x]); |
|
180 |
+ |
|
181 |
+ /* As a side-effect, we now know the maximum mask size, which |
|
182 |
+ * we'll use to generate our masks. */ |
|
183 |
+ /* Apply the fudge factor to this number too, since we must ensure |
|
184 |
+ * that enough masks are generated. */ |
|
185 |
+ *max_mask_size = apply_mask_fudge_factor(current_pass + 1); |
|
186 |
+} |
|
187 |
+ |
|
188 |
+static int query_formats(AVFilterContext *ctx) |
|
189 |
+{ |
|
190 |
+ enum PixelFormat pix_fmts[] = { PIX_FMT_YUV420P, PIX_FMT_NONE }; |
|
191 |
+ avfilter_set_common_pixel_formats(ctx, avfilter_make_format_list(pix_fmts)); |
|
192 |
+ return 0; |
|
193 |
+} |
|
194 |
+ |
|
195 |
+static int load_mask(uint8_t **mask, int *w, int *h, |
|
196 |
+ const char *filename, void *log_ctx) |
|
197 |
+{ |
|
198 |
+ int ret; |
|
199 |
+ enum PixelFormat pix_fmt; |
|
200 |
+ uint8_t *src_data[4], *gray_data[4]; |
|
201 |
+ int src_linesize[4], gray_linesize[4]; |
|
202 |
+ |
|
203 |
+ /* load image from file */ |
|
204 |
+ if ((ret = ff_load_image(src_data, src_linesize, w, h, &pix_fmt, filename, log_ctx)) < 0) |
|
205 |
+ return ret; |
|
206 |
+ |
|
207 |
+ /* convert the image to GRAY8 */ |
|
208 |
+ if ((ret = ff_scale_image(gray_data, gray_linesize, *w, *h, PIX_FMT_GRAY8, |
|
209 |
+ src_data, src_linesize, *w, *h, pix_fmt, |
|
210 |
+ log_ctx)) < 0) |
|
211 |
+ goto end; |
|
212 |
+ |
|
213 |
+ /* copy mask to a newly allocated array */ |
|
214 |
+ *mask = av_malloc(*w * *h); |
|
215 |
+ if (!*mask) |
|
216 |
+ ret = AVERROR(ENOMEM); |
|
217 |
+ av_image_copy_plane(*mask, *w, gray_data[0], gray_linesize[0], *w, *h); |
|
218 |
+ |
|
219 |
+end: |
|
220 |
+ av_free(src_data[0]); |
|
221 |
+ av_free(gray_data[0]); |
|
222 |
+ return ret; |
|
223 |
+} |
|
224 |
+ |
|
225 |
+/** |
|
226 |
+ * Generate a scaled down image with half width, height, and intensity. |
|
227 |
+ * |
|
228 |
+ * This function not only scales down an image, but halves the value |
|
229 |
+ * in each pixel too. The purpose of this is to produce a chroma |
|
230 |
+ * filter image out of a luma filter image. The pixel values store the |
|
231 |
+ * distance to the edge of the logo and halving the dimensions halves |
|
232 |
+ * the distance. This function rounds up, because a downwards rounding |
|
233 |
+ * error could cause the filter to fail, but an upwards rounding error |
|
234 |
+ * will only cause a minor amount of excess blur in the chroma planes. |
|
235 |
+ */ |
|
236 |
+static void generate_half_size_image(const uint8_t *src_data, int src_linesize, |
|
237 |
+ uint8_t *dst_data, int dst_linesize, |
|
238 |
+ int src_w, int src_h, |
|
239 |
+ int *max_mask_size) |
|
240 |
+{ |
|
241 |
+ int x, y; |
|
242 |
+ |
|
243 |
+ /* Copy over the image data, using the average of 4 pixels for to |
|
244 |
+ * calculate each downsampled pixel. */ |
|
245 |
+ for (y = 0; y < src_h/2; y++) { |
|
246 |
+ for (x = 0; x < src_w/2; x++) { |
|
247 |
+ /* Set the pixel if there exists a non-zero value in the |
|
248 |
+ * source pixels, else clear it. */ |
|
249 |
+ dst_data[(y * dst_linesize) + x] = |
|
250 |
+ src_data[((y << 1) * src_linesize) + (x << 1)] || |
|
251 |
+ src_data[((y << 1) * src_linesize) + (x << 1) + 1] || |
|
252 |
+ src_data[(((y << 1) + 1) * src_linesize) + (x << 1)] || |
|
253 |
+ src_data[(((y << 1) + 1) * src_linesize) + (x << 1) + 1]; |
|
254 |
+ dst_data[(y * dst_linesize) + x] = FFMIN(1, dst_data[(y * dst_linesize) + x]); |
|
255 |
+ } |
|
256 |
+ } |
|
257 |
+ |
|
258 |
+ convert_mask_to_strength_mask(dst_data, dst_linesize, |
|
259 |
+ src_w/2, src_h/2, 0, max_mask_size); |
|
260 |
+} |
|
261 |
+ |
|
262 |
+static av_cold int init(AVFilterContext *ctx, const char *args, void *opaque) |
|
263 |
+{ |
|
264 |
+ RemovelogoContext *removelogo = ctx->priv; |
|
265 |
+ int ***mask; |
|
266 |
+ int ret = 0; |
|
267 |
+ int a, b, c, w, h; |
|
268 |
+ int full_max_mask_size, half_max_mask_size; |
|
269 |
+ |
|
270 |
+ if (!args) { |
|
271 |
+ av_log(ctx, AV_LOG_ERROR, "An image file must be specified as argument\n"); |
|
272 |
+ return AVERROR(EINVAL); |
|
273 |
+ } |
|
274 |
+ |
|
275 |
+ /* Load our mask image. */ |
|
276 |
+ if ((ret = load_mask(&removelogo->full_mask_data, &w, &h, args, ctx)) < 0) |
|
277 |
+ return ret; |
|
278 |
+ removelogo->mask_w = w; |
|
279 |
+ removelogo->mask_h = h; |
|
280 |
+ |
|
281 |
+ convert_mask_to_strength_mask(removelogo->full_mask_data, w, w, h, |
|
282 |
+ 16, &full_max_mask_size); |
|
283 |
+ |
|
284 |
+ /* Create the scaled down mask image for the chroma planes. */ |
|
285 |
+ if (!(removelogo->half_mask_data = av_mallocz(w/2 * h/2))) |
|
286 |
+ return AVERROR(ENOMEM); |
|
287 |
+ generate_half_size_image(removelogo->full_mask_data, w, |
|
288 |
+ removelogo->half_mask_data, w/2, |
|
289 |
+ w, h, &half_max_mask_size); |
|
290 |
+ |
|
291 |
+ removelogo->max_mask_size = FFMAX(full_max_mask_size, half_max_mask_size); |
|
292 |
+ |
|
293 |
+ /* Create a circular mask for each size up to max_mask_size. When |
|
294 |
+ the filter is applied, the mask size is determined on a pixel |
|
295 |
+ by pixel basis, with pixels nearer the edge of the logo getting |
|
296 |
+ smaller mask sizes. */ |
|
297 |
+ mask = (int ***)av_malloc(sizeof(int **) * (removelogo->max_mask_size + 1)); |
|
298 |
+ if (!mask) |
|
299 |
+ return AVERROR(ENOMEM); |
|
300 |
+ |
|
301 |
+ for (a = 0; a <= removelogo->max_mask_size; a++) { |
|
302 |
+ mask[a] = (int **)av_malloc(sizeof(int *) * ((a * 2) + 1)); |
|
303 |
+ if (!mask[a]) |
|
304 |
+ return AVERROR(ENOMEM); |
|
305 |
+ for (b = -a; b <= a; b++) { |
|
306 |
+ mask[a][b + a] = (int *)av_malloc(sizeof(int) * ((a * 2) + 1)); |
|
307 |
+ if (!mask[a][b + a]) |
|
308 |
+ return AVERROR(ENOMEM); |
|
309 |
+ for (c = -a; c <= a; c++) { |
|
310 |
+ if ((b * b) + (c * c) <= (a * a)) /* Circular 0/1 mask. */ |
|
311 |
+ mask[a][b + a][c + a] = 1; |
|
312 |
+ else |
|
313 |
+ mask[a][b + a][c + a] = 0; |
|
314 |
+ } |
|
315 |
+ } |
|
316 |
+ } |
|
317 |
+ removelogo->mask = mask; |
|
318 |
+ |
|
319 |
+ /* Calculate our bounding rectangles, which determine in what |
|
320 |
+ * region the logo resides for faster processing. */ |
|
321 |
+ ff_calculate_bounding_box(&removelogo->full_mask_bbox, removelogo->full_mask_data, w, w, h, 0); |
|
322 |
+ ff_calculate_bounding_box(&removelogo->half_mask_bbox, removelogo->half_mask_data, w/2, w/2, h/2, 0); |
|
323 |
+ |
|
324 |
+#define SHOW_LOGO_INFO(mask_type) \ |
|
325 |
+ av_log(ctx, AV_LOG_INFO, #mask_type " x1:%d x2:%d y1:%d y2:%d max_mask_size:%d\n", \ |
|
326 |
+ removelogo->mask_type##_mask_bbox.x1, removelogo->mask_type##_mask_bbox.x2, \ |
|
327 |
+ removelogo->mask_type##_mask_bbox.y1, removelogo->mask_type##_mask_bbox.y2, \ |
|
328 |
+ mask_type##_max_mask_size); |
|
329 |
+ SHOW_LOGO_INFO(full); |
|
330 |
+ SHOW_LOGO_INFO(half); |
|
331 |
+ |
|
332 |
+ return 0; |
|
333 |
+} |
|
334 |
+ |
|
335 |
+static int config_props_input(AVFilterLink *inlink) |
|
336 |
+{ |
|
337 |
+ AVFilterContext *ctx = inlink->dst; |
|
338 |
+ RemovelogoContext *removelogo = ctx->priv; |
|
339 |
+ |
|
340 |
+ if (inlink->w != removelogo->mask_w || inlink->h != removelogo->mask_h) { |
|
341 |
+ av_log(ctx, AV_LOG_INFO, |
|
342 |
+ "Mask image size %dx%d does not match with the input video size %dx%d\n", |
|
343 |
+ removelogo->mask_w, removelogo->mask_h, inlink->w, inlink->h); |
|
344 |
+ return AVERROR(EINVAL); |
|
345 |
+ } |
|
346 |
+ |
|
347 |
+ return 0; |
|
348 |
+} |
|
349 |
+ |
|
350 |
+/** |
|
351 |
+ * Blur image. |
|
352 |
+ * |
|
353 |
+ * It takes a pixel that is inside the mask and blurs it. It does so |
|
354 |
+ * by finding the average of all the pixels within the mask and |
|
355 |
+ * outside of the mask. |
|
356 |
+ * |
|
357 |
+ * @param mask_data the mask plane to use for averaging |
|
358 |
+ * @param image_data the image plane to blur |
|
359 |
+ * @param w width of the image |
|
360 |
+ * @param h height of the image |
|
361 |
+ * @param x x-coordinate of the pixel to blur |
|
362 |
+ * @param y y-coordinate of the pixel to blur |
|
363 |
+ */ |
|
364 |
+static unsigned int blur_pixel(int ***mask, |
|
365 |
+ const uint8_t *mask_data, int mask_linesize, |
|
366 |
+ uint8_t *image_data, int image_linesize, |
|
367 |
+ int w, int h, int x, int y) |
|
368 |
+{ |
|
369 |
+ /* Mask size tells how large a circle to use. The radius is about |
|
370 |
+ * (slightly larger than) mask size. */ |
|
371 |
+ int mask_size; |
|
372 |
+ int start_posx, start_posy, end_posx, end_posy; |
|
373 |
+ int i, j; |
|
374 |
+ unsigned int accumulator = 0, divisor = 0; |
|
375 |
+ /* What pixel we are reading out of the circular blur mask. */ |
|
376 |
+ const uint8_t *image_read_position; |
|
377 |
+ /* What pixel we are reading out of the filter image. */ |
|
378 |
+ const uint8_t *mask_read_position; |
|
379 |
+ |
|
380 |
+ /* Prepare our bounding rectangle and clip it if need be. */ |
|
381 |
+ mask_size = mask_data[y * mask_linesize + x]; |
|
382 |
+ start_posx = FFMAX(0, x - mask_size); |
|
383 |
+ start_posy = FFMAX(0, y - mask_size); |
|
384 |
+ end_posx = FFMIN(w - 1, x + mask_size); |
|
385 |
+ end_posy = FFMIN(h - 1, y + mask_size); |
|
386 |
+ |
|
387 |
+ image_read_position = image_data + image_linesize * start_posy + start_posx; |
|
388 |
+ mask_read_position = mask_data + mask_linesize * start_posy + start_posx; |
|
389 |
+ |
|
390 |
+ for (j = start_posy; j <= end_posy; j++) { |
|
391 |
+ for (i = start_posx; i <= end_posx; i++) { |
|
392 |
+ /* Check if this pixel is in the mask or not. Only use the |
|
393 |
+ * pixel if it is not. */ |
|
394 |
+ if (!(*mask_read_position) && mask[mask_size][i - start_posx][j - start_posy]) { |
|
395 |
+ accumulator += *image_read_position; |
|
396 |
+ divisor++; |
|
397 |
+ } |
|
398 |
+ |
|
399 |
+ image_read_position++; |
|
400 |
+ mask_read_position++; |
|
401 |
+ } |
|
402 |
+ |
|
403 |
+ image_read_position += (image_linesize - ((end_posx + 1) - start_posx)); |
|
404 |
+ mask_read_position += (mask_linesize - ((end_posx + 1) - start_posx)); |
|
405 |
+ } |
|
406 |
+ |
|
407 |
+ /* If divisor is 0, it means that not a single pixel is outside of |
|
408 |
+ the logo, so we have no data. Else we need to normalise the |
|
409 |
+ data using the divisor. */ |
|
410 |
+ return divisor == 0 ? 255: |
|
411 |
+ (accumulator + (divisor / 2)) / divisor; /* divide, taking into account average rounding error */ |
|
412 |
+} |
|
413 |
+ |
|
414 |
+/** |
|
415 |
+ * Blur image plane using a mask. |
|
416 |
+ * |
|
417 |
+ * @param source The image to have it's logo removed. |
|
418 |
+ * @param destination Where the output image will be stored. |
|
419 |
+ * @param source_stride How far apart (in memory) two consecutive lines are. |
|
420 |
+ * @param destination Same as source_stride, but for the destination image. |
|
421 |
+ * @param width Width of the image. This is the same for source and destination. |
|
422 |
+ * @param height Height of the image. This is the same for source and destination. |
|
423 |
+ * @param is_image_direct If the image is direct, then source and destination are |
|
424 |
+ * the same and we can save a lot of time by not copying pixels that |
|
425 |
+ * haven't changed. |
|
426 |
+ * @param filter The image that stores the distance to the edge of the logo for |
|
427 |
+ * each pixel. |
|
428 |
+ * @param logo_start_x smallest x-coordinate that contains at least 1 logo pixel. |
|
429 |
+ * @param logo_start_y smallest y-coordinate that contains at least 1 logo pixel. |
|
430 |
+ * @param logo_end_x largest x-coordinate that contains at least 1 logo pixel. |
|
431 |
+ * @param logo_end_y largest y-coordinate that contains at least 1 logo pixel. |
|
432 |
+ * |
|
433 |
+ * This function processes an entire plane. Pixels outside of the logo are copied |
|
434 |
+ * to the output without change, and pixels inside the logo have the de-blurring |
|
435 |
+ * function applied. |
|
436 |
+ */ |
|
437 |
+static void blur_image(int ***mask, |
|
438 |
+ const uint8_t *src_data, int src_linesize, |
|
439 |
+ uint8_t *dst_data, int dst_linesize, |
|
440 |
+ const uint8_t *mask_data, int mask_linesize, |
|
441 |
+ int w, int h, int direct, |
|
442 |
+ FFBoundingBox *bbox) |
|
443 |
+{ |
|
444 |
+ int x, y; |
|
445 |
+ uint8_t *dst_line; |
|
446 |
+ const uint8_t *src_line; |
|
447 |
+ |
|
448 |
+ if (!direct) |
|
449 |
+ av_image_copy_plane(dst_data, dst_linesize, src_data, src_linesize, w, h); |
|
450 |
+ |
|
451 |
+ for (y = bbox->y1; y <= bbox->y2; y++) { |
|
452 |
+ src_line = src_data + src_linesize * y; |
|
453 |
+ dst_line = dst_data + dst_linesize * y; |
|
454 |
+ |
|
455 |
+ for (x = bbox->x1; x <= bbox->x2; x++) { |
|
456 |
+ if (mask_data[y * mask_linesize + x]) { |
|
457 |
+ /* Only process if we are in the mask. */ |
|
458 |
+ dst_line[x] = blur_pixel(mask, |
|
459 |
+ mask_data, mask_linesize, |
|
460 |
+ dst_data, dst_linesize, |
|
461 |
+ w, h, x, y); |
|
462 |
+ } else { |
|
463 |
+ /* Else just copy the data. */ |
|
464 |
+ if (!direct) |
|
465 |
+ dst_line[x] = src_line[x]; |
|
466 |
+ } |
|
467 |
+ } |
|
468 |
+ } |
|
469 |
+} |
|
470 |
+ |
|
471 |
+static void start_frame(AVFilterLink *inlink, AVFilterBufferRef *inpicref) |
|
472 |
+{ |
|
473 |
+ AVFilterLink *outlink = inlink->dst->outputs[0]; |
|
474 |
+ AVFilterBufferRef *outpicref; |
|
475 |
+ |
|
476 |
+ if (inpicref->perms & AV_PERM_PRESERVE) { |
|
477 |
+ outpicref = avfilter_get_video_buffer(outlink, AV_PERM_WRITE, |
|
478 |
+ outlink->w, outlink->h); |
|
479 |
+ avfilter_copy_buffer_ref_props(outpicref, inpicref); |
|
480 |
+ outpicref->video->w = outlink->w; |
|
481 |
+ outpicref->video->h = outlink->h; |
|
482 |
+ } else |
|
483 |
+ outpicref = inpicref; |
|
484 |
+ |
|
485 |
+ outlink->out_buf = outpicref; |
|
486 |
+ avfilter_start_frame(outlink, avfilter_ref_buffer(outpicref, ~0)); |
|
487 |
+} |
|
488 |
+ |
|
489 |
+static void end_frame(AVFilterLink *inlink) |
|
490 |
+{ |
|
491 |
+ RemovelogoContext *removelogo = inlink->dst->priv; |
|
492 |
+ AVFilterLink *outlink = inlink->dst->outputs[0]; |
|
493 |
+ AVFilterBufferRef *inpicref = inlink ->cur_buf; |
|
494 |
+ AVFilterBufferRef *outpicref = outlink->out_buf; |
|
495 |
+ int direct = inpicref == outpicref; |
|
496 |
+ |
|
497 |
+ blur_image(removelogo->mask, |
|
498 |
+ inpicref ->data[0], inpicref ->linesize[0], |
|
499 |
+ outpicref->data[0], outpicref->linesize[0], |
|
500 |
+ removelogo->full_mask_data, inlink->w, |
|
501 |
+ inlink->w, inlink->h, direct, &removelogo->full_mask_bbox); |
|
502 |
+ blur_image(removelogo->mask, |
|
503 |
+ inpicref ->data[1], inpicref ->linesize[1], |
|
504 |
+ outpicref->data[1], outpicref->linesize[1], |
|
505 |
+ removelogo->half_mask_data, inlink->w/2, |
|
506 |
+ inlink->w/2, inlink->h/2, direct, &removelogo->half_mask_bbox); |
|
507 |
+ blur_image(removelogo->mask, |
|
508 |
+ inpicref ->data[2], inpicref ->linesize[2], |
|
509 |
+ outpicref->data[2], outpicref->linesize[2], |
|
510 |
+ removelogo->half_mask_data, inlink->w/2, |
|
511 |
+ inlink->w/2, inlink->h/2, direct, &removelogo->half_mask_bbox); |
|
512 |
+ |
|
513 |
+ avfilter_draw_slice(outlink, 0, inlink->h, 1); |
|
514 |
+ avfilter_end_frame(outlink); |
|
515 |
+ avfilter_unref_buffer(inpicref); |
|
516 |
+ if (!direct) |
|
517 |
+ avfilter_unref_buffer(outpicref); |
|
518 |
+} |
|
519 |
+ |
|
520 |
+static void uninit(AVFilterContext *ctx) |
|
521 |
+{ |
|
522 |
+ RemovelogoContext *removelogo = ctx->priv; |
|
523 |
+ int a, b; |
|
524 |
+ |
|
525 |
+ av_freep(&removelogo->full_mask_data); |
|
526 |
+ av_freep(&removelogo->half_mask_data); |
|
527 |
+ |
|
528 |
+ if (removelogo->mask) { |
|
529 |
+ /* Loop through each mask. */ |
|
530 |
+ for (a = 0; a <= removelogo->max_mask_size; a++) { |
|
531 |
+ /* Loop through each scanline in a mask. */ |
|
532 |
+ for (b = -a; b <= a; b++) { |
|
533 |
+ av_free(removelogo->mask[a][b + a]); /* Free a scanline. */ |
|
534 |
+ } |
|
535 |
+ av_free(removelogo->mask[a]); |
|
536 |
+ } |
|
537 |
+ /* Free the array of pointers pointing to the masks. */ |
|
538 |
+ av_freep(&removelogo->mask); |
|
539 |
+ } |
|
540 |
+} |
|
541 |
+ |
|
542 |
+static void null_draw_slice(AVFilterLink *link, int y, int h, int slice_dir) { } |
|
543 |
+ |
|
544 |
+AVFilter avfilter_vf_removelogo = { |
|
545 |
+ .name = "removelogo", |
|
546 |
+ .description = NULL_IF_CONFIG_SMALL("Remove a TV logo based on a mask image."), |
|
547 |
+ .priv_size = sizeof(RemovelogoContext), |
|
548 |
+ .init = init, |
|
549 |
+ .uninit = uninit, |
|
550 |
+ .query_formats = query_formats, |
|
551 |
+ |
|
552 |
+ .inputs = (const AVFilterPad[]) { |
|
553 |
+ { .name = "default", |
|
554 |
+ .type = AVMEDIA_TYPE_VIDEO, |
|
555 |
+ .get_video_buffer = avfilter_null_get_video_buffer, |
|
556 |
+ .config_props = config_props_input, |
|
557 |
+ .draw_slice = null_draw_slice, |
|
558 |
+ .start_frame = start_frame, |
|
559 |
+ .end_frame = end_frame, |
|
560 |
+ .min_perms = AV_PERM_WRITE | AV_PERM_READ, |
|
561 |
+ .rej_perms = AV_PERM_PRESERVE }, |
|
562 |
+ { .name = NULL } |
|
563 |
+ }, |
|
564 |
+ .outputs = (const AVFilterPad[]) { |
|
565 |
+ { .name = "default", |
|
566 |
+ .type = AVMEDIA_TYPE_VIDEO, }, |
|
567 |
+ { .name = NULL } |
|
568 |
+ }, |
|
569 |
+}; |