FFmpeg
vf_blockdetect.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2021 Thilo Borgmann <thilo.borgmann _at_ mail.de>
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
23  * No-reference blockdetect filter
24  *
25  * Implementing:
26  * Remco Muijs and Ihor Kirenko: "A no-reference blocking artifact measure for adaptive video processing." 2005 13th European signal processing conference. IEEE, 2005.
27  * http://www.eurasip.org/Proceedings/Eusipco/Eusipco2005/defevent/papers/cr1042.pdf
28  *
29  * @author Thilo Borgmann <thilo.borgmann _at_ mail.de>
30  */
31 
32 #include "libavutil/imgutils.h"
33 #include "libavutil/opt.h"
34 #include "internal.h"
35 #include "video.h"
36 
37 typedef struct BLKContext {
38  const AVClass *class;
39 
40  int hsub, vsub;
41  int nb_planes;
42 
43  int period_min; // minimum period to search for
44  int period_max; // maximum period to search for
45  int planes; // number of planes to filter
46 
47  double block_total;
48  uint64_t nb_frames;
49 
50  float *gradients;
51 } BLKContext;
52 
53 #define OFFSET(x) offsetof(BLKContext, x)
54 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
55 static const AVOption blockdetect_options[] = {
56  { "period_min", "Minimum period to search for", OFFSET(period_min), AV_OPT_TYPE_INT, {.i64=3}, 2, 32, FLAGS},
57  { "period_max", "Maximum period to search for", OFFSET(period_max), AV_OPT_TYPE_INT, {.i64=24}, 2, 64, FLAGS},
58  { "planes", "set planes to filter", OFFSET(planes), AV_OPT_TYPE_INT, {.i64=1}, 0, 15, FLAGS },
59  { NULL }
60 };
61 
62 AVFILTER_DEFINE_CLASS(blockdetect);
63 
65 {
66  AVFilterContext *ctx = inlink->dst;
67  BLKContext *s = ctx->priv;
68  const int bufsize = inlink->w * inlink->h;
69  const AVPixFmtDescriptor *pix_desc;
70 
71  pix_desc = av_pix_fmt_desc_get(inlink->format);
72  s->hsub = pix_desc->log2_chroma_w;
73  s->vsub = pix_desc->log2_chroma_h;
74  s->nb_planes = av_pix_fmt_count_planes(inlink->format);
75 
76  s->gradients = av_calloc(bufsize, sizeof(*s->gradients));
77 
78  if (!s->gradients)
79  return AVERROR(ENOMEM);
80 
81  return 0;
82 }
83 
84 static float calculate_blockiness(BLKContext *s, int w, int h,
85  float *grad, int grad_linesize,
86  uint8_t* src, int src_linesize)
87 {
88  float block = 0.0f;
89  float nonblock = 0.0f;
90  int block_count = 0;
91  int nonblock_count = 0;
92  float ret = 0;
93 
94  // Calculate BS in horizontal and vertical directions according to (1)(2)(3).
95  // Also try to find integer pixel periods (grids) even for scaled images.
96  // In case of fractional periods, FFMAX of current and neighbor pixels
97  // can help improve the correlation with MQS.
98  // Skip linear correction term (4)(5), as it appears only valid for their own test samples.
99 
100  // horizontal blockiness (fixed width)
101  for (int j = 1; j < h; j++) {
102  for (int i = 3; i < w - 4; i++) {
103  float temp = 0.0f;
104  grad[j * grad_linesize + i] =
105  abs(src[j * src_linesize + i + 0] - src[j * src_linesize + i + 1]);
106  temp += abs(src[j * src_linesize + i + 1] - src[j * src_linesize + i + 2]);
107  temp += abs(src[j * src_linesize + i + 2] - src[j * src_linesize + i + 3]);
108  temp += abs(src[j * src_linesize + i + 3] - src[j * src_linesize + i + 4]);
109  temp += abs(src[j * src_linesize + i - 0] - src[j * src_linesize + i - 1]);
110  temp += abs(src[j * src_linesize + i - 1] - src[j * src_linesize + i - 2]);
111  temp += abs(src[j * src_linesize + i - 2] - src[j * src_linesize + i - 3]);
112  temp = FFMAX(1, temp);
113  grad[j * grad_linesize + i] /= temp;
114 
115  // use first row to store acculated results
116  grad[i] += grad[j * grad_linesize + i];
117  }
118  }
119 
120  // find horizontal period
121  for (int period = s->period_min; period < s->period_max + 1; period++) {
122  float temp;
123  block = 0;
124  nonblock = 0;
125  block_count = 0;
126  nonblock_count = 0;
127  for (int i = 3; i < w - 4; i++) {
128  if ((i % period) == (period - 1)) {
129  block += FFMAX(FFMAX(grad[i + 0], grad[i + 1]), grad[i - 1]);
130  block_count++;
131  } else {
132  nonblock += grad[i];
133  nonblock_count++;
134  }
135  }
136  if (block_count && nonblock_count) {
137  temp = (block / block_count) / (nonblock / nonblock_count);
138  ret = FFMAX(ret, temp);
139  }
140  }
141 
142  // vertical blockiness (fixed height)
143  block_count = 0;
144  for (int j = 3; j < h - 4; j++) {
145  for (int i = 1; i < w; i++) {
146  float temp = 0.0f;
147  grad[j * grad_linesize + i] =
148  abs(src[(j + 0) * src_linesize + i] - src[(j + 1) * src_linesize + i]);
149  temp += abs(src[(j + 1) * src_linesize + i] - src[(j + 2) * src_linesize + i]);
150  temp += abs(src[(j + 2) * src_linesize + i] - src[(j + 3) * src_linesize + i]);
151  temp += abs(src[(j + 3) * src_linesize + i] - src[(j + 4) * src_linesize + i]);
152  temp += abs(src[(j - 0) * src_linesize + i] - src[(j - 1) * src_linesize + i]);
153  temp += abs(src[(j - 1) * src_linesize + i] - src[(j - 2) * src_linesize + i]);
154  temp += abs(src[(j - 2) * src_linesize + i] - src[(j - 3) * src_linesize + i]);
155  temp = FFMAX(1, temp);
156  grad[j * grad_linesize + i] /= temp;
157 
158  // use first column to store accumulated results
159  grad[j * grad_linesize] += grad[j * grad_linesize + i];
160  }
161  }
162 
163  // find vertical period
164  for (int period = s->period_min; period < s->period_max + 1; period++) {
165  float temp;
166  block = 0;
167  nonblock = 0;
168  block_count = 0;
169  nonblock_count = 0;
170  for (int j = 3; j < h - 4; j++) {
171  if ((j % period) == (period - 1)) {
172  block += FFMAX(FFMAX(grad[(j + 0) * grad_linesize],
173  grad[(j + 1) * grad_linesize]),
174  grad[(j - 1) * grad_linesize]);
175  block_count++;
176  } else {
177  nonblock += grad[j * grad_linesize];
178  nonblock_count++;
179  }
180  }
181  if (block_count && nonblock_count) {
182  temp = (block / block_count) / (nonblock / nonblock_count);
183  ret = FFMAX(ret, temp);
184  }
185  }
186 
187  // return highest value of horz||vert
188  return ret;
189 }
190 
191 static void set_meta(AVDictionary **metadata, const char *key, float d)
192 {
193  char value[128];
194  snprintf(value, sizeof(value), "%f", d);
195  av_dict_set(metadata, key, value, 0);
196 }
197 
199 {
200  AVFilterContext *ctx = inlink->dst;
201  BLKContext *s = ctx->priv;
202  AVFilterLink *outlink = ctx->outputs[0];
203 
204  const int inw = inlink->w;
205  const int inh = inlink->h;
206 
207  float *gradients = s->gradients;
208 
209  float block = 0.0f;
210  int nplanes = 0;
211  AVDictionary **metadata;
212  metadata = &in->metadata;
213 
214  for (int plane = 0; plane < s->nb_planes; plane++) {
215  int hsub = plane == 1 || plane == 2 ? s->hsub : 0;
216  int vsub = plane == 1 || plane == 2 ? s->vsub : 0;
217  int w = AV_CEIL_RSHIFT(inw, hsub);
218  int h = AV_CEIL_RSHIFT(inh, vsub);
219 
220  if (!((1 << plane) & s->planes))
221  continue;
222 
223  nplanes++;
224 
225  block += calculate_blockiness(s, w, h, gradients, w, in->data[plane], in->linesize[plane]);
226  }
227 
228  if (nplanes)
229  block /= nplanes;
230 
231  s->block_total += block;
232 
233  // write stats
234  av_log(ctx, AV_LOG_VERBOSE, "block: %.7f\n", block);
235 
236  set_meta(metadata, "lavfi.block", block);
237 
238  s->nb_frames = inlink->frame_count_in;
239 
240  return ff_filter_frame(outlink, in);
241 }
242 
244 {
245  BLKContext *s = ctx->priv;
246 
247  if (s->nb_frames > 0) {
248  av_log(ctx, AV_LOG_INFO, "block mean: %.7f\n",
249  s->block_total / s->nb_frames);
250  }
251 
252  av_freep(&s->gradients);
253 }
254 
255 static const enum AVPixelFormat pix_fmts[] = {
265 };
266 
267 static const AVFilterPad blockdetect_inputs[] = {
268  {
269  .name = "default",
270  .type = AVMEDIA_TYPE_VIDEO,
271  .config_props = blockdetect_config_input,
272  .filter_frame = blockdetect_filter_frame,
273  },
274 };
275 
277  .name = "blockdetect",
278  .description = NULL_IF_CONFIG_SMALL("Blockdetect filter."),
279  .priv_size = sizeof(BLKContext),
284  .priv_class = &blockdetect_class,
286 };
BLKContext::gradients
float * gradients
Definition: vf_blockdetect.c:50
AVPixelFormat
AVPixelFormat
Pixel format.
Definition: pixfmt.h:64
AVERROR
Filter the word “frame” indicates either a video frame or a group of audio as stored in an AVFrame structure Format for each input and each output the list of supported formats For video that means pixel format For audio that means channel sample they are references to shared objects When the negotiation mechanism computes the intersection of the formats supported at each end of a all references to both lists are replaced with a reference to the intersection And when a single format is eventually chosen for a link amongst the remaining all references to the list are updated That means that if a filter requires that its input and output have the same format amongst a supported all it has to do is use a reference to the same list of formats query_formats can leave some formats unset and return AVERROR(EAGAIN) to cause the negotiation mechanism toagain later. That can be used by filters with complex requirements to use the format negotiated on one link to set the formats supported on another. Frame references ownership and permissions
opt.h
ff_filter_frame
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
Definition: avfilter.c:978
av_pix_fmt_desc_get
const AVPixFmtDescriptor * av_pix_fmt_desc_get(enum AVPixelFormat pix_fmt)
Definition: pixdesc.c:2964
BLKContext::period_max
int period_max
Definition: vf_blockdetect.c:44
FILTER_PIXFMTS_ARRAY
#define FILTER_PIXFMTS_ARRAY(array)
Definition: internal.h:172
inlink
The exact code depends on how similar the blocks are and how related they are to the and needs to apply these operations to the correct inlink or outlink if there are several Macros are available to factor that when no extra processing is inlink
Definition: filter_design.txt:212
BLKContext::period_min
int period_min
Definition: vf_blockdetect.c:43
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:340
w
uint8_t w
Definition: llviddspenc.c:38
pix_fmts
static enum AVPixelFormat pix_fmts[]
Definition: vf_blockdetect.c:255
AVOption
AVOption.
Definition: opt.h:251
AV_LOG_VERBOSE
#define AV_LOG_VERBOSE
Detailed information.
Definition: log.h:196
blockdetect_inputs
static const AVFilterPad blockdetect_inputs[]
Definition: vf_blockdetect.c:267
AV_PIX_FMT_YUV440P
@ AV_PIX_FMT_YUV440P
planar YUV 4:4:0 (1 Cr & Cb sample per 1x2 Y samples)
Definition: pixfmt.h:99
AVDictionary
Definition: dict.c:34
FFMAX
#define FFMAX(a, b)
Definition: macros.h:47
AVFilter::name
const char * name
Filter name.
Definition: avfilter.h:170
blockdetect_uninit
static av_cold void blockdetect_uninit(AVFilterContext *ctx)
Definition: vf_blockdetect.c:243
video.h
BLKContext::planes
int planes
Definition: vf_blockdetect.c:45
AVFrame::data
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
Definition: frame.h:361
hsub
static void hsub(htype *dst, const htype *src, int bins)
Definition: vf_median.c:73
av_pix_fmt_count_planes
int av_pix_fmt_count_planes(enum AVPixelFormat pix_fmt)
Definition: pixdesc.c:3004
ff_vf_blockdetect
const AVFilter ff_vf_blockdetect
Definition: vf_blockdetect.c:276
AV_PIX_FMT_GBRAP
@ AV_PIX_FMT_GBRAP
planar GBRA 4:4:4:4 32bpp
Definition: pixfmt.h:205
BLKContext
Definition: vf_blockdetect.c:37
set_meta
static void set_meta(AVDictionary **metadata, const char *key, float d)
Definition: vf_blockdetect.c:191
AVFilterPad
A filter pad used for either input or output.
Definition: internal.h:47
AV_PIX_FMT_YUVJ411P
@ AV_PIX_FMT_YUVJ411P
planar YUV 4:1:1, 12bpp, (1 Cr & Cb sample per 4x1 Y samples) full scale (JPEG), deprecated in favor ...
Definition: pixfmt.h:276
av_cold
#define av_cold
Definition: attributes.h:90
ff_video_default_filterpad
const AVFilterPad ff_video_default_filterpad[1]
An AVFilterPad array whose only entry has name "default" and is of type AVMEDIA_TYPE_VIDEO.
Definition: video.c:36
AV_PIX_FMT_YUVJ422P
@ AV_PIX_FMT_YUVJ422P
planar YUV 4:2:2, 16bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV422P and setting col...
Definition: pixfmt.h:79
s
#define s(width, name)
Definition: cbs_vp9.c:198
AV_PIX_FMT_YUVA420P
@ AV_PIX_FMT_YUVA420P
planar YUV 4:2:0, 20bpp, (1 Cr & Cb sample per 2x2 Y & A samples)
Definition: pixfmt.h:101
AV_CEIL_RSHIFT
#define AV_CEIL_RSHIFT(a, b)
Definition: common.h:51
ctx
AVFormatContext * ctx
Definition: movenc.c:48
AVPixFmtDescriptor::log2_chroma_w
uint8_t log2_chroma_w
Amount to shift the luma width right to find the chroma width.
Definition: pixdesc.h:80
AV_PIX_FMT_YUV420P
@ AV_PIX_FMT_YUV420P
planar YUV 4:2:0, 12bpp, (1 Cr & Cb sample per 2x2 Y samples)
Definition: pixfmt.h:66
key
const char * key
Definition: hwcontext_opencl.c:174
FILTER_INPUTS
#define FILTER_INPUTS(array)
Definition: internal.h:192
AV_PIX_FMT_YUVJ444P
@ AV_PIX_FMT_YUVJ444P
planar YUV 4:4:4, 24bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV444P and setting col...
Definition: pixfmt.h:80
OFFSET
#define OFFSET(x)
Definition: vf_blockdetect.c:53
BLKContext::nb_planes
int nb_planes
Definition: vf_blockdetect.c:41
AVClass
Describe the class of an AVClass context structure.
Definition: log.h:66
NULL
#define NULL
Definition: coverity.c:32
FLAGS
#define FLAGS
Definition: vf_blockdetect.c:54
period
it s the only field you need to keep assuming you have a context There is some magic you don t need to care about around this just let it vf default minimum maximum flags name is the option keep it simple and lowercase description are in without period
Definition: writing_filters.txt:89
AV_PIX_FMT_YUVJ420P
@ AV_PIX_FMT_YUVJ420P
planar YUV 4:2:0, 12bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV420P and setting col...
Definition: pixfmt.h:78
abs
#define abs(x)
Definition: cuda_runtime.h:35
AV_PIX_FMT_GRAY8
@ AV_PIX_FMT_GRAY8
Y , 8bpp.
Definition: pixfmt.h:74
calculate_blockiness
static float calculate_blockiness(BLKContext *s, int w, int h, float *grad, int grad_linesize, uint8_t *src, int src_linesize)
Definition: vf_blockdetect.c:84
BLKContext::hsub
int hsub
Definition: vf_blockdetect.c:40
NULL_IF_CONFIG_SMALL
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification.
Definition: internal.h:106
AV_PIX_FMT_YUVA444P
@ AV_PIX_FMT_YUVA444P
planar YUV 4:4:4 32bpp, (1 Cr & Cb sample per 1x1 Y & A samples)
Definition: pixfmt.h:167
AV_LOG_INFO
#define AV_LOG_INFO
Standard information.
Definition: log.h:191
internal.h
i
#define i(width, name, range_min, range_max)
Definition: cbs_h2645.c:255
BLKContext::block_total
double block_total
Definition: vf_blockdetect.c:47
planes
static const struct @363 planes[]
value
it s the only field you need to keep assuming you have a context There is some magic you don t need to care about around this just let it vf default value
Definition: writing_filters.txt:86
AV_PIX_FMT_YUVJ440P
@ AV_PIX_FMT_YUVJ440P
planar YUV 4:4:0 full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV440P and setting color_range
Definition: pixfmt.h:100
blockdetect_filter_frame
static int blockdetect_filter_frame(AVFilterLink *inlink, AVFrame *in)
Definition: vf_blockdetect.c:198
AVFilterPad::name
const char * name
Pad name.
Definition: internal.h:53
BLKContext::vsub
int vsub
Definition: vf_blockdetect.c:40
av_calloc
void * av_calloc(size_t nmemb, size_t size)
Definition: mem.c:262
AVFilter
Filter definition.
Definition: avfilter.h:166
ret
ret
Definition: filter_design.txt:187
AV_PIX_FMT_NONE
@ AV_PIX_FMT_NONE
Definition: pixfmt.h:65
AV_OPT_TYPE_INT
@ AV_OPT_TYPE_INT
Definition: opt.h:225
AVFrame::metadata
AVDictionary * metadata
metadata.
Definition: frame.h:708
AVFILTER_FLAG_METADATA_ONLY
#define AVFILTER_FLAG_METADATA_ONLY
The filter is a "metadata" filter - it does not modify the frame data in any way.
Definition: avfilter.h:133
temp
else temp
Definition: vf_mcdeint.c:263
AV_PIX_FMT_YUV444P
@ AV_PIX_FMT_YUV444P
planar YUV 4:4:4, 24bpp, (1 Cr & Cb sample per 1x1 Y samples)
Definition: pixfmt.h:71
AVFilterContext
An instance of a filter.
Definition: avfilter.h:397
AV_PIX_FMT_GBRP
@ AV_PIX_FMT_GBRP
planar GBR 4:4:4 24bpp
Definition: pixfmt.h:158
blockdetect_options
static const AVOption blockdetect_options[]
Definition: vf_blockdetect.c:55
AVMEDIA_TYPE_VIDEO
@ AVMEDIA_TYPE_VIDEO
Definition: avutil.h:201
AV_PIX_FMT_YUV422P
@ AV_PIX_FMT_YUV422P
planar YUV 4:2:2, 16bpp, (1 Cr & Cb sample per 2x1 Y samples)
Definition: pixfmt.h:70
AVFILTER_DEFINE_CLASS
AVFILTER_DEFINE_CLASS(blockdetect)
AVPixFmtDescriptor
Descriptor that unambiguously describes how the bits of a pixel are stored in the up to 4 data planes...
Definition: pixdesc.h:69
FILTER_OUTPUTS
#define FILTER_OUTPUTS(array)
Definition: internal.h:193
av_freep
#define av_freep(p)
Definition: tableprint_vlc.h:34
av_dict_set
int av_dict_set(AVDictionary **pm, const char *key, const char *value, int flags)
Set the given entry in *pm, overwriting an existing entry.
Definition: dict.c:88
src
INIT_CLIP pixel * src
Definition: h264pred_template.c:418
AV_PIX_FMT_YUV411P
@ AV_PIX_FMT_YUV411P
planar YUV 4:1:1, 12bpp, (1 Cr & Cb sample per 4x1 Y samples)
Definition: pixfmt.h:73
d
d
Definition: ffmpeg_filter.c:368
imgutils.h
AVFrame::linesize
int linesize[AV_NUM_DATA_POINTERS]
For video, a positive or negative value, which is typically indicating the size in bytes of each pict...
Definition: frame.h:385
AV_PIX_FMT_YUV410P
@ AV_PIX_FMT_YUV410P
planar YUV 4:1:0, 9bpp, (1 Cr & Cb sample per 4x4 Y samples)
Definition: pixfmt.h:72
block
The exact code depends on how similar the blocks are and how related they are to the block
Definition: filter_design.txt:207
av_log
#define av_log(a,...)
Definition: tableprint_vlc.h:27
uninit
static av_cold int uninit(AVCodecContext *avctx)
Definition: crystalhd.c:285
h
h
Definition: vp9dsp_template.c:2038
BLKContext::nb_frames
uint64_t nb_frames
Definition: vf_blockdetect.c:48
snprintf
#define snprintf
Definition: snprintf.h:34
AVPixFmtDescriptor::log2_chroma_h
uint8_t log2_chroma_h
Amount to shift the luma height right to find the chroma height.
Definition: pixdesc.h:89
AV_PIX_FMT_YUVA422P
@ AV_PIX_FMT_YUVA422P
planar YUV 4:2:2 24bpp, (1 Cr & Cb sample per 2x1 Y & A samples)
Definition: pixfmt.h:166
blockdetect_config_input
static int blockdetect_config_input(AVFilterLink *inlink)
Definition: vf_blockdetect.c:64