Go to the documentation of this file.
67 int jobnr,
int nb_jobs);
71 #define OFFSET(x) offsetof(SSIMContext, x)
72 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
96 const uint8_t *ref8, ptrdiff_t ref_stride,
99 const uint16_t *main16 = (
const uint16_t *)main8;
100 const uint16_t *ref16 = (
const uint16_t *)ref8;
106 for (z = 0; z <
width; z++) {
107 uint64_t s1 = 0, s2 = 0,
ss = 0, s12 = 0;
109 for (y = 0; y < 4; y++) {
110 for (x = 0; x < 4; x++) {
111 unsigned a = main16[x + y * main_stride];
112 unsigned b = ref16[x + y * ref_stride];
132 const uint8_t *
ref, ptrdiff_t ref_stride,
133 int (*sums)[4],
int width)
137 for (z = 0; z <
width; z++) {
138 uint32_t s1 = 0, s2 = 0,
ss = 0, s12 = 0;
140 for (y = 0; y < 4; y++) {
141 for (x = 0; x < 4; x++) {
142 int a =
main[x + y * main_stride];
143 int b =
ref[x + y * ref_stride];
172 int64_t covar = fs12 * 64 - fs1 * fs2;
174 return (
float)(2 * fs1 * fs2 + ssim_c1) * (
float)(2 * covar + ssim_c2)
175 / ((
float)(fs1 * fs1 + fs2 * fs2 + ssim_c1) * (
float)(
vars + ssim_c2));
180 static const int ssim_c1 = (int)(.01*.01*255*255*64 + .5);
181 static const int ssim_c2 = (int)(.03*.03*255*255*64*63 + .5);
187 int vars = fss * 64 - fs1 * fs1 - fs2 * fs2;
188 int covar = fs12 * 64 - fs1 * fs2;
190 return (
float)(2 * fs1 * fs2 + ssim_c1) * (
float)(2 * covar + ssim_c2)
191 / ((
float)(fs1 * fs1 + fs2 * fs2 + ssim_c1) * (
float)(
vars + ssim_c2));
199 ssim +=
ssim_end1x(sum0[
i][0] + sum0[
i + 1][0] + sum1[
i][0] + sum1[
i + 1][0],
200 sum0[
i][1] + sum0[
i + 1][1] + sum1[
i][1] + sum1[
i + 1][1],
201 sum0[
i][2] + sum0[
i + 1][2] + sum1[
i][2] + sum1[
i + 1][2],
202 sum0[
i][3] + sum0[
i + 1][3] + sum1[
i][3] + sum1[
i + 1][3],
212 ssim +=
ssim_end1(sum0[
i][0] + sum0[
i + 1][0] + sum1[
i][0] + sum1[
i + 1][0],
213 sum0[
i][1] + sum0[
i + 1][1] + sum1[
i][1] + sum1[
i + 1][1],
214 sum0[
i][2] + sum0[
i + 1][2] + sum1[
i][2] + sum1[
i + 1][2],
215 sum0[
i][3] + sum0[
i + 1][3] + sum1[
i][3] + sum1[
i + 1][3]);
219 #define SUM_LEN(w) (((w) >> 2) + 3)
236 int jobnr,
int nb_jobs)
239 double *score = td->
score[jobnr];
245 const uint8_t *ref_data = td->
ref_data[
c];
261 for (
int y = ystart; y <
slice_end; y++) {
262 for (; z <= y; z++) {
263 FFSWAP(
void*, sum0, sum1);
265 &ref_data[4 * z * ref_stride], ref_stride,
279 int jobnr,
int nb_jobs)
282 double *score = td->
score[jobnr];
288 const uint8_t *ref_data = td->
ref_data[
c];
298 int (*sum0)[4] =
temp;
304 for (
int y = ystart; y <
slice_end; y++) {
305 for (; z <= y; z++) {
306 FFSWAP(
void*, sum0, sum1);
307 dsp->
ssim_4x4_line(&main_data[4 * z * main_stride], main_stride,
308 &ref_data[4 * z * ref_stride], ref_stride,
312 ssim += dsp->
ssim_end_line((
const int (*)[4])sum0, (
const int (*)[4])sum1,
width - 1);
332 double c[4] = {0}, ssimv = 0.0;
339 if (
ctx->is_disabled || !
ref)
341 metadata = &
master->metadata;
351 for (
int n = 0; n <
s->nb_components; n++) {
360 if (
master->color_range !=
ref->color_range) {
362 "frames use different color ranges (%s != %s)\n",
368 FFMIN((
s->planeheight[1] + 3) >> 2,
s->nb_threads));
370 for (
i = 0;
i <
s->nb_components;
i++) {
371 for (
int j = 0; j <
s->nb_threads; j++)
372 c[
i] +=
s->score[j][
i];
373 c[
i] =
c[
i] / (((
s->planewidth[
i] >> 2) - 1) * ((
s->planeheight[
i] >> 2) - 1));
376 for (
i = 0;
i <
s->nb_components;
i++) {
377 ssimv +=
s->coefs[
i] *
c[
i];
381 for (
i = 0;
i <
s->nb_components;
i++) {
382 int cidx =
s->is_rgb ?
s->rgba_map[
i] :
i;
383 set_meta(metadata,
"lavfi.ssim.",
s->comps[
i],
c[cidx]);
385 s->ssim_total += ssimv;
387 set_meta(metadata,
"lavfi.ssim.All", 0, ssimv);
391 fprintf(
s->stats_file,
"n:%"PRId64
" ",
s->nb_frames);
393 for (
i = 0;
i <
s->nb_components;
i++) {
394 int cidx =
s->is_rgb ?
s->rgba_map[
i] :
i;
395 fprintf(
s->stats_file,
"%c:%f ",
s->comps[
i],
c[cidx]);
398 fprintf(
s->stats_file,
"All:%f (%f)\n", ssimv,
ssim_db(ssimv, 1.0));
408 if (
s->stats_file_str) {
409 if (!strcmp(
s->stats_file_str,
"-")) {
410 s->stats_file = stdout;
413 if (!
s->stats_file) {
434 #define PF(suf) AV_PIX_FMT_YUV420##suf, AV_PIX_FMT_YUV422##suf, AV_PIX_FMT_YUV444##suf, AV_PIX_FMT_GBR##suf
447 s->nb_components =
desc->nb_components;
449 if (
ctx->inputs[0]->w !=
ctx->inputs[1]->w ||
450 ctx->inputs[0]->h !=
ctx->inputs[1]->h) {
456 s->comps[0] =
s->is_rgb ?
'R' :
'Y';
457 s->comps[1] =
s->is_rgb ?
'G' :
'U';
458 s->comps[2] =
s->is_rgb ?
'B' :
'V';
462 s->planeheight[0] =
s->planeheight[3] =
inlink->h;
464 s->planewidth[0] =
s->planewidth[3] =
inlink->w;
465 for (
int i = 0;
i <
s->nb_components;
i++)
466 sum +=
s->planeheight[
i] *
s->planewidth[
i];
467 for (
int i = 0;
i <
s->nb_components;
i++)
468 s->coefs[
i] = (
double)
s->planeheight[
i] *
s->planewidth[
i] / sum;
474 for (
int t = 0; t <
s->nb_threads; t++) {
479 s->max = (1 <<
desc->comp[0].depth) - 1;
492 for (
int t = 0; t <
s->nb_threads; t++) {
493 s->score[t] =
av_calloc(
s->nb_components,
sizeof(*
s->score[0]));
513 outlink->
w = mainlink->
w;
514 outlink->
h = mainlink->
h;
526 av_log(
ctx,
AV_LOG_WARNING,
"not matching timebases found between first input: %d/%d and second input %d/%d, results may be incorrect!\n",
528 ctx->inputs[1]->time_base.num,
ctx->inputs[1]->time_base.den);
543 if (
s->nb_frames > 0) {
546 for (
int i = 0;
i <
s->nb_components;
i++) {
547 int c =
s->is_rgb ?
s->rgba_map[
i] :
i;
548 av_strlcatf(buf,
sizeof(buf),
" %c:%f (%f)",
s->comps[
i],
s->ssim[
c] /
s->nb_frames,
552 s->ssim_total /
s->nb_frames,
ssim_db(
s->ssim_total,
s->nb_frames));
557 if (
s->stats_file &&
s->stats_file != stdout)
558 fclose(
s->stats_file);
560 for (
int t = 0; t <
s->nb_threads &&
s->score; t++)
564 for (
int t = 0; t <
s->nb_threads &&
s->temp; t++)
591 .preinit = ssim_framesync_preinit,
596 .priv_class = &ssim_class,
int ff_framesync_configure(FFFrameSync *fs)
Configure a frame sync structure.
void ff_ssim_init_x86(SSIMDSPContext *dsp)
#define AV_LOG_WARNING
Something somehow does not look correct.
AVPixelFormat
Pixel format.
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
#define FILTER_PIXFMTS_ARRAY(array)
void ff_framesync_uninit(FFFrameSync *fs)
Free all memory currently allocated.
static void comp(unsigned char *dst, ptrdiff_t dst_stride, unsigned char *src, ptrdiff_t src_stride, int add)
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
const AVPixFmtDescriptor * av_pix_fmt_desc_get(enum AVPixelFormat pix_fmt)
static float ssim_end1x(int64_t s1, int64_t s2, int64_t ss, int64_t s12, int max)
static int activate(AVFilterContext *ctx)
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
static int config_input_ref(AVFilterLink *inlink)
#define FILTER_INPUTS(array)
This structure describes decoded (raw) audio or video data.
@ AV_PIX_FMT_YUV440P
planar YUV 4:4:0 (1 Cr & Cb sample per 1x2 Y samples)
const char * name
Filter name.
A link between two filters.
size_t av_strlcatf(char *dst, size_t size, const char *fmt,...)
Link properties exposed to filter code, but not external callers.
static int slice_end(AVCodecContext *avctx, AVFrame *pict, int *got_output)
Handle slice ends.
static const AVOption ssim_options[]
const h264_weight_func weight
#define AV_PIX_FMT_GRAY16
#define ss(width, name, subs,...)
A filter pad used for either input or output.
@ AV_PIX_FMT_YUVJ411P
planar YUV 4:1:1, 12bpp, (1 Cr & Cb sample per 4x1 Y samples) full scale (JPEG), deprecated in favor ...
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
@ AV_PIX_FMT_YUVJ422P
planar YUV 4:2:2, 16bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV422P and setting col...
static float ssim_endn_16bit(const int64_t(*sum0)[4], const int64_t(*sum1)[4], int width, int max)
void(* ssim_4x4_line)(const uint8_t *buf, ptrdiff_t buf_stride, const uint8_t *ref, ptrdiff_t ref_stride, int(*sums)[4], int w)
static double ssim_endn_8bit(const int(*sum0)[4], const int(*sum1)[4], int width)
#define AV_CEIL_RSHIFT(a, b)
static const AVFilterPad ssim_inputs[]
AVRational sample_aspect_ratio
agreed upon sample aspect ratio
#define AV_PIX_FMT_GRAY14
@ AV_PIX_FMT_YUV420P
planar YUV 4:2:0, 12bpp, (1 Cr & Cb sample per 2x2 Y samples)
#define FILTER_OUTPUTS(array)
@ AV_PIX_FMT_YUVJ444P
planar YUV 4:4:4, 24bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV444P and setting col...
static void ssim_4x4xn_8bit(const uint8_t *main, ptrdiff_t main_stride, const uint8_t *ref, ptrdiff_t ref_stride, int(*sums)[4], int width)
#define AV_PIX_FMT_GRAY10
const char * av_color_range_name(enum AVColorRange range)
Describe the class of an AVClass context structure.
static __device__ float fabs(float a)
#define fs(width, name, subs,...)
static const uint8_t vars[2][12]
static int ssim_plane(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
@ AV_PIX_FMT_YUVJ420P
planar YUV 4:2:0, 12bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV420P and setting col...
FRAMESYNC_DEFINE_CLASS(ssim, SSIMContext, fs)
@ AV_PIX_FMT_GRAY8
Y , 8bpp.
Undefined Behavior In the C some operations are like signed integer dereferencing freed accessing outside allocated Undefined Behavior must not occur in a C it is not safe even if the output of undefined operations is unused The unsafety may seem nit picking but Optimizing compilers have in fact optimized code on the assumption that no undefined Behavior occurs Optimizing code based on wrong assumptions can and has in some cases lead to effects beyond the output of computations The signed integer overflow problem in speed critical code Code which is highly optimized and works with signed integers sometimes has the problem that often the output of the computation does not c
static FilterLink * ff_filter_link(AVFilterLink *link)
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification.
int ff_framesync_init_dualinput(FFFrameSync *fs, AVFilterContext *parent)
Initialize a frame sync structure for dualinput.
static double ssim_db(double ssim, double weight)
#define av_err2str(errnum)
Convenience macro, the return value should be used only directly in function arguments but never stan...
const AVFilter ff_vf_ssim
static av_cold int init(AVFilterContext *ctx)
AVFilterContext * src
source filter
The reader does not expect b to be semantically here and if the code is changed by maybe adding a a division or other the signedness will almost certainly be mistaken To avoid this confusion a new type was SUINT is the C unsigned type but it holds a signed int to use the same example SUINT a
#define AV_LOG_INFO
Standard information.
static av_cold void uninit(AVFilterContext *ctx)
static const AVFilterPad ssim_outputs[]
static int do_ssim(FFFrameSync *fs)
static void ssim_4x4xn_16bit(const uint8_t *main8, ptrdiff_t main_stride, const uint8_t *ref8, ptrdiff_t ref_stride, int64_t(*sums)[4], int width)
static float ssim_end1(int s1, int s2, int ss, int s12)
#define i(width, name, range_min, range_max)
int w
agreed upon image width
double(* ssim_end_line)(const int(*sum0)[4], const int(*sum1)[4], int w)
int ff_filter_get_nb_threads(AVFilterContext *ctx)
Get number of threads for current filter instance.
Used for passing data between threads.
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
@ AV_PIX_FMT_YUVJ440P
planar YUV 4:4:0 full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV440P and setting color_range
const char * name
Pad name.
FILE * avpriv_fopen_utf8(const char *path, const char *mode)
Open a file using a UTF-8 filename.
void * av_calloc(size_t nmemb, size_t size)
static int slice_start(SliceContext *sc, VVCContext *s, VVCFrameContext *fc, const CodedBitstreamUnit *unit, const int is_first_slice)
static int av_cmp_q(AVRational a, AVRational b)
Compare two rationals.
static int config_output(AVFilterLink *outlink)
#define FFSWAP(type, a, b)
int(* ssim_plane)(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
static enum AVPixelFormat pix_fmts[]
int h
agreed upon image height
int ff_filter_execute(AVFilterContext *ctx, avfilter_action_func *func, void *arg, int *ret, int nb_jobs)
#define AVFILTER_FLAG_METADATA_ONLY
The filter is a "metadata" filter - it does not modify the frame data in any way.
static int ref[MAX_W *MAX_W]
static void set_meta(AVDictionary **metadata, const char *key, char comp, float d)
AVRational time_base
Define the time base used by the PTS of the frames/samples which will pass through this link.
@ AV_PIX_FMT_YUV444P
planar YUV 4:4:4, 24bpp, (1 Cr & Cb sample per 1x1 Y samples)
@ AV_PIX_FMT_GBRP
planar GBR 4:4:4 24bpp
#define AVFILTER_FLAG_SLICE_THREADS
The filter supports multithreading by splitting frames into multiple parts and processing them concur...
@ AV_PIX_FMT_YUV422P
planar YUV 4:2:2, 16bpp, (1 Cr & Cb sample per 2x1 Y samples)
Descriptor that unambiguously describes how the bits of a pixel are stored in the up to 4 data planes...
int av_dict_set(AVDictionary **pm, const char *key, const char *value, int flags)
Set the given entry in *pm, overwriting an existing entry.
@ AV_PIX_FMT_YUV411P
planar YUV 4:1:1, 12bpp, (1 Cr & Cb sample per 4x1 Y samples)
int ff_fill_rgba_map(uint8_t *rgba_map, enum AVPixelFormat pix_fmt)
#define AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL
Same as AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC, except that the filter will have its filter_frame() c...
@ AV_PIX_FMT_YUV410P
planar YUV 4:1:0, 9bpp, (1 Cr & Cb sample per 4x4 Y samples)
static int ssim_plane_16bit(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
AVRational frame_rate
Frame rate of the stream on the link, or 1/0 if unknown or variable.
int ff_framesync_activate(FFFrameSync *fs)
Examine the frames in the filter's input and try to produce output.
int ff_framesync_dualinput_get(FFFrameSync *fs, AVFrame **f0, AVFrame **f1)
@ AV_OPT_TYPE_STRING
Underlying C type is a uint8_t* that is either NULL or points to a C string allocated with the av_mal...
#define AV_PIX_FMT_GRAY12
const uint8_t * ref_data[4]
const uint8_t * main_data[4]