Go to the documentation of this file.
37 static const uint8_t
NNEDI_XDIM[] = { 8, 16, 32, 48, 8, 16, 32 };
38 static const uint8_t
NNEDI_YDIM[] = { 6, 6, 6, 6, 4, 4, 4 };
39 static const uint16_t
NNEDI_NNS[] = { 16, 32, 64, 128, 256 };
106 int src_stride,
int dst_stride,
109 int src_stride,
int dst_stride,
112 const void *
src, ptrdiff_t src_stride,
117 #define OFFSET(x) offsetof(NNEDIContext, x)
118 #define RFLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_RUNTIME_PARAM
119 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
125 {
"interlaced",
"only deinterlace frames marked as interlaced", 0,
AV_OPT_TYPE_CONST, {.i64=1}, 0, 0,
RFLAGS, .unit =
"deint" },
127 {
"af",
"use frame flags, both fields", 0,
AV_OPT_TYPE_CONST, {.i64=-2}, 0, 0,
RFLAGS, .unit =
"field" },
128 {
"a",
"use frame flags, single field", 0,
AV_OPT_TYPE_CONST, {.i64=-1}, 0, 0,
RFLAGS, .unit =
"field" },
131 {
"tf",
"use both fields, top first", 0,
AV_OPT_TYPE_CONST, {.i64=2}, 0, 0,
RFLAGS, .unit =
"field" },
132 {
"bf",
"use both fields, bottom first", 0,
AV_OPT_TYPE_CONST, {.i64=3}, 0, 0,
RFLAGS, .unit =
"field" },
134 {
"nsize",
"set size of local neighborhood around each pixel, used by the predictor neural network",
OFFSET(nsize),
AV_OPT_TYPE_INT, {.i64=6}, 0, 6,
RFLAGS, .unit =
"nsize" },
142 {
"nns",
"set number of neurons in predictor neural network",
OFFSET(nnsparam),
AV_OPT_TYPE_INT, {.i64=1}, 0, 4,
RFLAGS, .unit =
"nns" },
151 {
"etype",
"set which set of weights to use in the predictor",
OFFSET(etype),
AV_OPT_TYPE_INT, {.i64=0}, 0, 1,
RFLAGS, .unit =
"etype" },
152 {
"a",
"weights trained to minimize absolute error", 0,
AV_OPT_TYPE_CONST, {.i64=0}, 0, 0,
RFLAGS, .unit =
"etype" },
153 {
"abs",
"weights trained to minimize absolute error", 0,
AV_OPT_TYPE_CONST, {.i64=0}, 0, 0,
RFLAGS, .unit =
"etype" },
154 {
"s",
"weights trained to minimize squared error", 0,
AV_OPT_TYPE_CONST, {.i64=1}, 0, 0,
RFLAGS, .unit =
"etype" },
155 {
"mse",
"weights trained to minimize squared error", 0,
AV_OPT_TYPE_CONST, {.i64=1}, 0, 0,
RFLAGS, .unit =
"etype" },
173 outlink->
w =
ctx->inputs[0]->w;
174 outlink->
h =
ctx->inputs[0]->h;
176 if (
s->field == -2 ||
s->field > 1) {
216 sum =
s->fdsp->scalarproduct_float(kernel,
input, n);
225 return x / (1.0f +
fabsf(x));
235 const void *
src, ptrdiff_t src_stride,
236 uint8_t *prescreen,
int N,
240 const float *src_p =
src;
243 const float *
window = src_p - 2 * src_stride - 5;
245 for (
int j = 0; j <
N; j++) {
249 for (
int i = 0;
i < 4;
i++)
250 memcpy(
input +
i * 12,
window +
i * src_stride + j, 12 *
sizeof(
float));
253 for (
int n = 0; n < 4; n++)
258 for (
int n = 0; n < 4; n++)
263 for (
int n = 0; n < 4; n++)
271 const void *
src, ptrdiff_t src_stride,
272 uint8_t *prescreen,
int N,
276 const float *src_p =
src;
279 const float *
window = src_p - 2 * src_stride - 6;
281 for (
int j = 0; j <
N; j += 4) {
285 for (
int i = 0;
i < 4;
i++)
286 memcpy(
input +
i * 16,
window +
i * src_stride + j, 16 *
sizeof(
float));
288 for (
int n = 0; n < 4; n++)
292 for (
int n = 0; n < 4; n++)
295 for (
int n = 0; n < 4; n++)
296 prescreen[j + n] =
state[n + 4] > 0.
f;
302 return nn * model->
nsize;
330 float *buf,
float mstd[4],
338 for (
int i = 0;
i < model->
ydim;
i++) {
339 memcpy(buf,
src, model->
xdim *
sizeof(
float));
341 for (
int j = 0; j < model->
xdim; j++) {
352 mstd[0] = sum *
scale;
355 tmp = sum_sq *
scale - mstd[0] * mstd[0];
356 if (
tmp < FLT_EPSILON) {
361 mstd[2] = 1.0f / mstd[1];
376 static void wae5(
const float *softmax,
const float *el,
377 int n,
float mstd[4])
379 float vsum = 0.0f, wsum = 0.0f;
381 for (
int i = 0;
i < n;
i++) {
387 mstd[3] += (5.0f * vsum) / wsum * mstd[1] + mstd[0];
393 const void *
src, ptrdiff_t src_stride,
void *dst,
394 const uint8_t *prescreen,
int N,
398 const float *src_p =
src;
402 const float *
window = src_p - (model->
ydim / 2) * src_stride - (model->
xdim / 2 - 1);
403 const int filter_size = model->
nsize;
404 const int nns = model->
nns;
406 for (
int i = 0;
i <
N;
i++) {
408 float activation[256 * 2];
418 for (
int nn = 0; nn < nns; nn++)
421 for (
int nn = 0; nn < nns; nn++)
425 wae5(activation, activation + nns, nns, mstd);
428 for (
int nn = 0; nn < nns; nn++)
431 for (
int nn = 0; nn < nns; nn++)
435 wae5(activation, activation + nns, nns, mstd);
438 dst_p[
i] = mstd[3] * (use_q2 ? 0.5f : 1.f);
443 int src_stride,
int dst_stride,
446 for (
int y = 0; y <
height; y++) {
447 for (
int x = 0; x < 32; x++)
448 dst[-x - 1] =
src[x];
450 for (
int x = 0; x <
width; x++)
453 for (
int x = 0; x < 32; x++)
462 int src_stride,
int dst_stride,
465 const uint16_t *
src = (
const uint16_t *)srcp;
469 for (
int y = 0; y <
height; y++) {
470 for (
int x = 0; x < 32; x++)
473 for (
int x = 0; x <
width; x++)
476 for (
int x = 0; x < 32; x++)
485 int src_stride,
int dst_stride,
489 for (
int y = 0; y <
height; y++) {
490 for (
int x = 0; x <
width; x++)
499 int src_stride,
int dst_stride,
503 uint16_t *dst = (uint16_t *)dstp;
507 for (
int y = 0; y <
height; y++) {
508 for (
int x = 0; x <
width; x++)
517 void *dst,
const uint8_t *prescreen,
int n)
519 const float *src_p =
src;
521 const float *
window = src_p - 2 * src_stride;
523 for (
int i = 0;
i < n;
i++) {
529 accum += (-3.0f / 32.0f) *
window[0 * src_stride +
i];
530 accum += (19.0f / 32.0f) *
window[1 * src_stride +
i];
531 accum += (19.0f / 32.0f) *
window[2 * src_stride +
i];
532 accum += (-3.0f / 32.0f) *
window[3 * src_stride +
i];
543 const float in_scale =
s->in_scale;
544 const float out_scale =
s->out_scale;
545 const int depth =
s->depth;
551 for (
int p = 0; p <
s->nb_planes; p++) {
552 const int height =
s->planeheight[p];
553 const int width =
s->planewidth[p];
556 const uint8_t *src_data = in->
data[p];
557 uint8_t *dst_data =
out->data[p];
559 const int src_linesize = in->
linesize[p];
560 const int dst_linesize =
out->linesize[p];
561 uint8_t *prescreen_buf =
s->prescreen_buf[jobnr];
562 float *srcbuf =
s->input_buf[jobnr];
563 const int srcbuf_stride =
width + 64;
564 float *dstbuf =
s->output_buf[jobnr];
565 const int dstbuf_stride =
width;
568 const uint8_t *in_line;
572 if (!(
s->process_plane & (1 << p))) {
581 in_line = src_data + (y_out * src_linesize);
582 out_line = dst_data + (y_out * dst_linesize);
585 memcpy(out_line, in_line,
s->linesize[p]);
587 in_line += src_linesize * 2;
588 out_line += dst_linesize * 2;
593 s->read(src_data +
FFMAX(y_out - 5, tff) * src_linesize,
595 src_linesize * 2, srcbuf_stride,
597 srcbuf += srcbuf_stride;
599 s->read(src_data +
FFMAX(y_out - 3, tff) * src_linesize,
601 src_linesize * 2, srcbuf_stride,
603 srcbuf += srcbuf_stride;
605 s->read(src_data +
FFMAX(y_out - 1, tff) * src_linesize,
607 src_linesize * 2, srcbuf_stride,
609 srcbuf += srcbuf_stride;
611 in_line = src_data +
FFMIN(y_out + 1,
height - 1 - !tff) * src_linesize;
612 out_line = dst_data + (y_out * dst_linesize);
614 s->read(in_line, srcbuf + 32, src_linesize * 2, srcbuf_stride,
615 width, slice_height - last_slice, in_scale);
617 y_out += (slice_height - last_slice) * 2;
619 s->read(src_data +
FFMIN(y_out + 1,
height - 1 - !tff) * src_linesize,
620 srcbuf + 32 + srcbuf_stride * (slice_height - last_slice),
621 src_linesize * 2, srcbuf_stride,
624 s->read(src_data +
FFMIN(y_out + 3,
height - 1 - !tff) * src_linesize,
625 srcbuf + 32 + srcbuf_stride * (slice_height + 1 - last_slice),
626 src_linesize * 2, srcbuf_stride,
629 s->read(src_data +
FFMIN(y_out + 5,
height - 1 - !tff) * src_linesize,
630 srcbuf + 32 + srcbuf_stride * (slice_height + 2 - last_slice),
631 src_linesize * 2, srcbuf_stride,
636 s->prescreen[
s->pscrn > 1](
ctx, srcbuf + (y / 2) * srcbuf_stride + 32,
637 srcbuf_stride, prescreen_buf,
width,
638 &
s->prescreener[
s->pscrn - 1]);
641 srcbuf + (y / 2) * srcbuf_stride + 32,
643 dstbuf + (y / 2) * dstbuf_stride,
644 prescreen_buf,
width,
645 &
s->coeffs[
s->etype][
s->nnsparam][
s->nsize],
s->qual == 2);
650 dstbuf + (y / 2) * dstbuf_stride,
651 prescreen_buf,
width);
654 s->write(dstbuf, out_line, dstbuf_stride, dst_linesize * 2,
655 width, slice_height, depth, out_scale);
671 #if FF_API_INTERLACED_FRAME
680 FFMIN(
s->planeheight[1] / 2,
s->nb_threads));
682 if (
s->field == -2 ||
s->field > 1)
683 s->field_n = !
s->field_n;
706 s->pts =
s->prev->pts * 2;
708 if (
ret < 0 || (
s->field > -2 &&
s->field < 2)) {
714 s->pts =
s->prev->pts + in->
pts;
740 ctx->outputs[0]->time_base);
744 }
else if (
ret < 0) {
753 memcpy(dst, *
data, n *
sizeof(
float));
768 int filter_size = nns * xdim * ydim;
772 data =
av_calloc(filter_size + bias_size, 4 *
sizeof(
float));
779 coeffs->
nsize = xdim * ydim;
800 copy_weights(&
s->prescreener[0].kernel_l0[0][0], 4 * 48, &bdata);
803 copy_weights(&
s->prescreener[0].kernel_l1[0][0], 4 * 4, &bdata);
806 copy_weights(&
s->prescreener[0].kernel_l2[0][0], 4 * 8, &bdata);
809 for (
int i = 0;
i < 3;
i++) {
811 float kernel_l0_shuffled[4 * 64];
812 float kernel_l1_shuffled[4 * 4];
820 for (
int n = 0; n < 4; n++) {
821 for (
int k = 0; k < 64; k++)
822 data->kernel_l0[n][k] = kernel_l0_shuffled[(k / 8) * 32 + n * 8 + k % 8];
823 for (
int k = 0; k < 4; k++)
824 data->kernel_l1[n][k] = kernel_l1_shuffled[k * 4 + n];
828 for (
int m = 0; m < 2; m++) {
830 for (
int i = 0;
i < 5;
i++) {
834 for (
int j = 0; j < 7; j++) {
838 const int filter_size = xdim * ydim;
884 for (
int n = 0; n < 4; n++) {
893 for (
int n = 0; n < 4; n++) {
902 const int filter_size = model->
nsize;
903 const int nns = model->
nns;
904 const float scale = 1.f / nns;
906 double softmax_means[256];
907 double elliott_means[256];
908 double mean_filter[48 * 6] = { 0 };
912 for (
int nn = 0; nn < nns; nn++) {
913 softmax_means[nn] =
mean(model->
softmax_q1 + nn * filter_size, filter_size);
914 elliott_means[nn] =
mean(model->
elliott_q1 + nn * filter_size, filter_size);
916 for (
int k = 0; k < filter_size; k++)
917 mean_filter[k] += model->
softmax_q1[nn * filter_size + k] - softmax_means[nn];
920 for (
int k = 0; k < filter_size; k++)
921 mean_filter[k] *=
scale;
925 for (
int nn = 0; nn < nns; nn++) {
926 for (
int k = 0; k < filter_size; k++) {
927 model->
softmax_q1[nn * filter_size + k] -= softmax_means[nn] + mean_filter[k];
928 model->
elliott_q1[nn * filter_size + k] -= elliott_means[nn];
934 memset(mean_filter, 0,
sizeof(mean_filter));
936 for (
int nn = 0; nn < nns; nn++) {
937 softmax_means[nn] =
mean(model->
softmax_q2 + nn * filter_size, filter_size);
938 elliott_means[nn] =
mean(model->
elliott_q2 + nn * filter_size, filter_size);
940 for (
int k = 0; k < filter_size; k++) {
941 mean_filter[k] += model->
softmax_q2[nn * filter_size + k] - softmax_means[nn];
945 for (
int k = 0; k < filter_size; k++)
946 mean_filter[k] *=
scale;
950 for (
int nn = 0; nn < nns; nn++) {
951 for (
int k = 0; k < filter_size; k++) {
952 model->
softmax_q2[nn * filter_size + k] -= softmax_means[nn] + mean_filter[k];
953 model->
elliott_q2[nn * filter_size + k] -= elliott_means[nn];
963 FILE *weights_file =
NULL;
975 if (fseek(weights_file, 0, SEEK_END)) {
977 fclose(weights_file);
981 weights_size = ftell(weights_file);
983 if (weights_size == -1) {
984 fclose(weights_file);
988 fclose(weights_file);
993 if (fseek(weights_file, 0, SEEK_SET)) {
994 fclose(weights_file);
1001 fclose(weights_file);
1007 fclose(weights_file);
1013 fclose(weights_file);
1037 s->depth =
desc->comp[0].depth;
1044 s->planewidth[0] =
s->planewidth[3] =
inlink->w;
1046 s->planeheight[0] =
s->planeheight[3] =
inlink->h;
1048 s->half = ((1 << 8) - 1) / 2.f;
1049 s->out_scale = 1 << (
s->depth - 8);
1050 s->in_scale = 1.f /
s->out_scale;
1071 for (
int i = 0;
i < 2;
i++) {
1072 for (
int j = 0; j < 5; j++) {
1073 for (
int k = 0; k < 7; k++)
1078 s->input_size = (
s->planewidth[0] + 64) * (
s->planeheight[0] + 6);
1079 s->input_buf =
av_calloc(
s->nb_threads,
sizeof(*
s->input_buf));
1083 for (
int i = 0;
i <
s->nb_threads;
i++) {
1084 s->input_buf[
i] =
av_calloc(
s->input_size,
sizeof(**
s->input_buf));
1085 if (!
s->input_buf[
i])
1089 s->output_buf =
av_calloc(
s->nb_threads,
sizeof(*
s->output_buf));
1093 for (
int i = 0;
i <
s->nb_threads;
i++) {
1094 s->output_buf[
i] =
av_calloc(
s->input_size,
sizeof(**
s->output_buf));
1095 if (!
s->output_buf[
i])
1099 s->prescreen_buf =
av_calloc(
s->nb_threads,
sizeof(*
s->prescreen_buf));
1100 if (!
s->prescreen_buf)
1103 for (
int i = 0;
i <
s->nb_threads;
i++) {
1104 s->prescreen_buf[
i] =
av_calloc(
s->planewidth[0],
sizeof(**
s->prescreen_buf));
1105 if (!
s->prescreen_buf[
i])
1116 for (
int i = 0;
i <
s->nb_threads &&
s->prescreen_buf;
i++)
1121 for (
int i = 0;
i <
s->nb_threads &&
s->input_buf;
i++)
1126 for (
int i = 0;
i <
s->nb_threads &&
s->output_buf;
i++)
1132 for (
int i = 0;
i < 2;
i++) {
1133 for (
int j = 0; j < 5; j++) {
1134 for (
int k = 0; k < 7; k++) {
1163 .description =
NULL_IF_CONFIG_SMALL(
"Apply neural network edge directed interpolation intra-only deinterlacer."),
1165 .priv_class = &nnedi_class,
AVFrame * ff_get_video_buffer(AVFilterLink *link, int w, int h)
Request a picture buffer with a specific set of permissions.
#define AV_PIX_FMT_YUVA422P16
#define AV_PIX_FMT_GBRAP16
#define FF_ENABLE_DEPRECATION_WARNINGS
static float dot_dsp(const NNEDIContext *const s, const float *kernel, const float *input, int n, float scale, float bias)
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
static int get_frame(AVFilterContext *ctx, int is_second)
static void subtract_mean_predictor(PredictorCoefficients *model)
#define FILTER_PIXFMTS_ARRAY(array)
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)
#define AVERROR_EOF
End of file.
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 const float * softmax_q2_filter(int nn, const PredictorCoefficients *const model)
void av_frame_free(AVFrame **frame)
Free the frame and any dynamically allocated objects in it, e.g.
#define AV_PIX_FMT_YUVA422P9
#define FILTER_INPUTS(array)
This structure describes decoded (raw) audio or video data.
static av_always_inline av_const unsigned av_clip_uintp2_c(int a, int p)
Clip a signed integer to an unsigned power of two range.
int64_t pts
Presentation timestamp in time_base units (time when frame should be shown to user).
#define AV_PIX_FMT_YUVA420P16
#define AV_PIX_FMT_YUVA420P10
static void read_bytes(const uint8_t *src, float *dst, int src_stride, int dst_stride, int width, int height, float scale)
static uint8_t half(int a, int b)
#define AV_PIX_FMT_YUV420P10
int ff_request_frame(AVFilterLink *link)
Request an input frame from the filter at the other end of the link.
static void process_old(AVFilterContext *ctx, const void *src, ptrdiff_t src_stride, uint8_t *prescreen, int N, const PrescreenerCoefficients *const m_data)
void(* prescreen[2])(AVFilterContext *ctx, const void *src, ptrdiff_t src_stride, uint8_t *prescreen, int N, const PrescreenerCoefficients *const coeffs)
AVFILTER_DEFINE_CLASS(nnedi)
static av_cold void uninit(AVFilterContext *ctx)
@ AV_PIX_FMT_YUV440P
planar YUV 4:4:0 (1 Cr & Cb sample per 1x2 Y samples)
int flags
Frame flags, a combination of AV_FRAME_FLAGS.
const char * name
Filter name.
A link between two filters.
#define AV_PIX_FMT_YUVA422P10
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
void av_image_copy_plane(uint8_t *dst, int dst_linesize, const uint8_t *src, int src_linesize, int bytewidth, int height)
Copy image plane from src to dst.
Link properties exposed to filter code, but not external callers.
#define AV_FRAME_FLAG_TOP_FIELD_FIRST
A flag to mark frames where the top field is displayed first if the content is interlaced.
static float elliott(float x)
int av_pix_fmt_count_planes(enum AVPixelFormat pix_fmt)
#define AV_PIX_FMT_YUVA420P9
static SDL_Window * window
static const float * elliott_q2_filter(int nn, const PredictorCoefficients *const model)
#define AV_PIX_FMT_GBRP14
static int slice_end(AVCodecContext *avctx, AVFrame *pict, int *got_output)
Handle slice ends.
@ AV_PIX_FMT_GBRAP
planar GBRA 4:4:4:4 32bpp
#define AV_PIX_FMT_GBRP10
#define AV_PIX_FMT_YUVA444P16
#define AV_PIX_FMT_YUV422P9
const AVFilter ff_vf_nnedi
static double val(void *priv, double ch)
static const uint8_t NNEDI_XDIM[]
#define AV_PIX_FMT_GRAY16
static __device__ float fabsf(float a)
A filter pad used for either input or output.
static void write_words(const float *src, uint8_t *dstp, int src_stride, int dst_stride, int width, int height, int depth, float scale)
static float softmax_exp(float x)
#define AV_PIX_FMT_YUV444P10
@ AV_PIX_FMT_YUVJ411P
planar YUV 4:1:1, 12bpp, (1 Cr & Cb sample per 4x1 Y samples) full scale (JPEG), deprecated in favor ...
attribute_deprecated int interlaced_frame
The content of the picture is interlaced.
static int filter_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
static const AVFilterPad inputs[]
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
#define AV_PIX_FMT_YUV422P16
static float * allocate(float **ptr, int size)
@ AV_PIX_FMT_YUVJ422P
planar YUV 4:2:2, 16bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV422P and setting col...
#define AV_PIX_FMT_GBRAP10
static const uint16_t NNEDI_NNS[]
int av_image_fill_linesizes(int linesizes[4], enum AVPixelFormat pix_fmt, int width)
Fill plane linesizes for an image with pixel format pix_fmt and width width.
#define AV_PIX_FMT_GBRAP12
@ AV_PIX_FMT_YUVA420P
planar YUV 4:2:0, 20bpp, (1 Cr & Cb sample per 2x2 Y & A samples)
#define AV_PIX_FMT_YUV444P16
#define AV_CEIL_RSHIFT(a, b)
float kernel_l0[4][16 *4]
static int request_frame(AVFilterLink *link)
#define AV_PIX_FMT_YUVA444P12
#define AV_PIX_FMT_YUV420P9
#define AV_PIX_FMT_YUV420P16
#define AV_PIX_FMT_GRAY14
AVFrame * av_frame_clone(const AVFrame *src)
Create a new frame that references the same data as src.
int64_t av_rescale_q(int64_t a, AVRational bq, AVRational cq)
Rescale a 64-bit integer by 2 rational numbers.
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 field
@ AV_PIX_FMT_YUV420P
planar YUV 4:2:0, 12bpp, (1 Cr & Cb sample per 2x2 Y samples)
#define FILTER_OUTPUTS(array)
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 link
@ AV_PIX_FMT_YUVJ444P
planar YUV 4:4:4, 24bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV444P and setting col...
#define AV_PIX_FMT_GRAY10
#define AV_PIX_FMT_GBRP16
Describe the class of an AVClass context structure.
static const float * softmax_q1_filter(int nn, const PredictorCoefficients *const model)
static void interpolation(const void *src, ptrdiff_t src_stride, void *dst, const uint8_t *prescreen, int n)
#define LOCAL_ALIGNED_32(t, v,...)
int av_frame_copy_props(AVFrame *dst, const AVFrame *src)
Copy only "metadata" fields from src to dst.
PredictorCoefficients coeffs[2][5][7]
static int bias(int x, int c)
static void subtract_mean_old(PrescreenerCoefficients *coeffs, float half)
Rational number (pair of numerator and denominator).
@ AV_PIX_FMT_YUVJ420P
planar YUV 4:2:0, 12bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV420P and setting col...
static void transform_elliott(float *input, int size)
#define AV_PIX_FMT_YUV440P10
void(* write)(const float *src, uint8_t *dst, int src_stride, int dst_stride, int width, int height, int depth, float scale)
static const uint8_t NNEDI_YDIM[]
static __device__ float sqrtf(float a)
static const AVOption nnedi_options[]
#define AV_PIX_FMT_YUV422P10
@ AV_PIX_FMT_GRAY8
Y , 8bpp.
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.
#define DECLARE_ALIGNED(n, t, v)
#define AV_PIX_FMT_YUV422P12
static void wae5(const float *softmax, const float *el, int n, float mstd[4])
#define AV_PIX_FMT_YUV444P12
static const float * elliott_q1_filter(int nn, const PredictorCoefficients *const model)
AVFilterContext * src
source filter
int ff_filter_process_command(AVFilterContext *ctx, const char *cmd, const char *arg, char *res, int res_len, int flags)
Generic processing of user supplied commands that are set in the same way as the filter options.
static void predictor(AVFilterContext *ctx, const void *src, ptrdiff_t src_stride, void *dst, const uint8_t *prescreen, int N, const PredictorCoefficients *const model, int use_q2)
@ AV_PIX_FMT_YUVA444P
planar YUV 4:4:4 32bpp, (1 Cr & Cb sample per 1x1 Y & A samples)
#define AV_PIX_FMT_YUVA444P10
static void transform_softmax_exp(float *input, int size)
and forward the test the status of outputs and forward it to the corresponding return FFERROR_NOT_READY If the filters stores internally one or a few frame for some input
static const AVFilterPad outputs[]
static av_cold int init(AVFilterContext *ctx)
static int filter_offset(int nn, const PredictorCoefficients *const model)
#define i(width, name, range_min, range_max)
int w
agreed upon image width
#define AV_PIX_FMT_GBRP12
int ff_filter_get_nb_threads(AVFilterContext *ctx)
Get number of threads for current filter instance.
@ AV_PIX_FMT_YUVJ440P
planar YUV 4:4:0 full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV440P and setting color_range
static enum AVPixelFormat pix_fmts[]
static av_always_inline AVRational av_inv_q(AVRational q)
Invert a rational.
const char * name
Pad name.
FILE * avpriv_fopen_utf8(const char *path, const char *mode)
Open a file using a UTF-8 filename.
#define AV_FRAME_FLAG_INTERLACED
A flag to mark frames whose content is interlaced.
static void write_bytes(const float *src, uint8_t *dst, int src_stride, int dst_stride, int width, int height, int depth, float scale)
void * av_calloc(size_t nmemb, size_t size)
void(* read)(const uint8_t *src, float *dst, int src_stride, int dst_stride, int width, int height, float scale)
#define AV_PIX_FMT_YUV444P9
static void subtract_mean_new(PrescreenerCoefficients *coeffs, float half)
#define AV_PIX_FMT_YUVA444P9
static void gather_input(const float *src, ptrdiff_t src_stride, float *buf, float mstd[4], const PredictorCoefficients *const model)
#define AV_PIX_FMT_YUV420P12
#define AV_PIX_FMT_YUV422P14
static int allocate_model(PredictorCoefficients *coeffs, int xdim, int ydim, int nns)
int h
agreed upon image height
int ff_filter_execute(AVFilterContext *ctx, avfilter_action_func *func, void *arg, int *ret, int nb_jobs)
static const size_t NNEDI_WEIGHTS_SIZE
#define AV_PIX_FMT_YUVA422P12
@ AV_OPT_TYPE_INT
Underlying C type is int.
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
static void process_new(AVFilterContext *ctx, const void *src, ptrdiff_t src_stride, uint8_t *prescreen, int N, const PrescreenerCoefficients *const m_data)
static int slice_start(SliceContext *sc, VVCContext *s, VVCFrameContext *fc, const CodedBitstreamUnit *unit, const int is_first_slice)
static float mean(const float *input, int size)
AVRational time_base
Define the time base used by the PTS of the frames/samples which will pass through this link.
static void read_words(const uint8_t *srcp, float *dst, int src_stride, int dst_stride, int width, int height, float scale)
static void transform(float *input, int size, float mean, float half)
AVRational av_mul_q(AVRational b, AVRational c)
Multiply two rationals.
@ AV_PIX_FMT_YUV444P
planar YUV 4:4:4, 24bpp, (1 Cr & Cb sample per 1x1 Y samples)
#define FF_DISABLE_DEPRECATION_WARNINGS
@ 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...
static void scale(int *out, const int *in, const int w, const int h, const int shift)
static int read_weights(AVFilterContext *ctx, const float *bdata)
av_cold AVFloatDSPContext * avpriv_float_dsp_alloc(int bit_exact)
Allocate a float DSP context.
@ AV_PIX_FMT_YUV411P
planar YUV 4:1:1, 12bpp, (1 Cr & Cb sample per 4x1 Y samples)
static int config_input(AVFilterLink *inlink)
#define AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL
Same as AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC, except that the filter will have its filter_frame() c...
int linesize[AV_NUM_DATA_POINTERS]
For video, a positive or negative value, which is typically indicating the size in bytes of each pict...
@ AV_PIX_FMT_YUV410P
planar YUV 4:1:0, 9bpp, (1 Cr & Cb sample per 4x4 Y samples)
#define AVERROR_INVALIDDATA
Invalid data found when processing input.
static void copy_weights(float *dst, int n, const float **data)
#define AV_PIX_FMT_YUV440P12
AVRational frame_rate
Frame rate of the stream on the link, or 1/0 if unknown or variable.
#define AV_PIX_FMT_YUV444P14
@ 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
@ AV_OPT_TYPE_CONST
Special option type for declaring named constants.
@ AV_PIX_FMT_YUVA422P
planar YUV 4:2:2 24bpp, (1 Cr & Cb sample per 2x1 Y & A samples)
#define AV_PIX_FMT_YUV420P14
PrescreenerCoefficients prescreener[4]
static int config_output(AVFilterLink *outlink)