October 2021 - ffmpeg-cvslog - lists.ffmpeg.org

libaom: Dont use aom_codec_av1_dx_algo.
by Matt Oliver 30 Oct '21

30 Oct '21

ffmpeg | branch: master | Matt Oliver <protogonoi(a)gmail.com> | Wed Oct 20 07:02:39 2021 +1100| [d92fdc714496d43234733c315894abe0beeb3529] | committer: Matt Oliver libaom: Dont use aom_codec_av1_dx_algo. This fixes linking errors where variables cannot be correctly linked in from an external shared library such as with msvc (requires dllimport which is not used by libaom). Instead just call the function that returns the same variable. Signed-off-by: Matt Oliver <protogonoi(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d92fdc714496d4323… --- libavcodec/libaomdec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavcodec/libaomdec.c b/libavcodec/libaomdec.c index 75ecc08970..d6b822fda2 100644 --- a/libavcodec/libaomdec.c +++ b/libavcodec/libaomdec.c @@ -241,7 +241,7 @@ static av_cold int aom_free(AVCodecContext *avctx) static av_cold int av1_init(AVCodecContext *avctx) { - return aom_init(avctx, &aom_codec_av1_dx_algo); + return aom_init(avctx, aom_codec_av1_dx()); } const AVCodec ff_libaom_av1_decoder = {

1 0

doc/filters: fix typo for hsvkey/hsvhold
by Paul B Mahol 29 Oct '21

29 Oct '21

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Fri Oct 29 23:24:44 2021 +0200| [4f8494fe5a4307e8661e4d66231f488d644a0524] | committer: Paul B Mahol doc/filters: fix typo for hsvkey/hsvhold > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=4f8494fe5a4307e86… --- doc/filters.texi | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/doc/filters.texi b/doc/filters.texi index e5e25ef05a..177f0774fc 100644 --- a/doc/filters.texi +++ b/doc/filters.texi @@ -13735,15 +13735,15 @@ The filter accepts the following options: @table @option @item hue -Set the hue value which will be used if color difference calculation. +Set the hue value which will be used in color difference calculation. Allowed range is from -360 to 360. Default value is 0. @item sat -Set the saturation value which will be used if color difference calculation. +Set the saturation value which will be used in color difference calculation. Allowed range is from -1 to 1. Default value is 0. @item val -Set the value which will be used if color difference calculation. +Set the value which will be used in color difference calculation. Allowed range is from -1 to 1. Default value is 0. @item similarity @@ -13773,15 +13773,15 @@ The filter accepts the following options: @table @option @item hue -Set the hue value which will be used if color difference calculation. +Set the hue value which will be used in color difference calculation. Allowed range is from -360 to 360. Default value is 0. @item sat -Set the saturation value which will be used if color difference calculation. +Set the saturation value which will be used in color difference calculation. Allowed range is from -1 to 1. Default value is 0. @item val -Set the value which will be used if color difference calculation. +Set the value which will be used in color difference calculation. Allowed range is from -1 to 1. Default value is 0. @item similarity

1 0

avfilter: add huesaturation filter
by Paul B Mahol 29 Oct '21

29 Oct '21

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Tue Sep 28 09:25:43 2021 +0200| [8164fe113910f83e0989d5d151c744895a710710] | committer: Paul B Mahol avfilter: add huesaturation filter > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=8164fe113910f83e0… --- Changelog | 1 + doc/filters.texi | 56 +++++ libavfilter/Makefile | 1 + libavfilter/allfilters.c | 1 + libavfilter/version.h | 2 +- libavfilter/vf_huesaturation.c | 480 +++++++++++++++++++++++++++++++++++++++++ 6 files changed, 540 insertions(+), 1 deletion(-) diff --git a/Changelog b/Changelog index d806b96253..765ec82915 100644 --- a/Changelog +++ b/Changelog @@ -29,6 +29,7 @@ version <next>: - limitdiff video filter - xcorrelate video filter - varblur video filter +- huesaturation video filter version 4.4: diff --git a/doc/filters.texi b/doc/filters.texi index ae6cb3aafd..e5e25ef05a 100644 --- a/doc/filters.texi +++ b/doc/filters.texi @@ -13908,6 +13908,62 @@ If the specified expression is not valid, it is kept at its current value. @end table +@section huesaturation +Apply hue-saturation-intensity adjustments to input video stream. + +This filter operates in RGB colorspace. + +This filter accepts the following options: + +@table @option +@item hue +Set the hue shift in degrees to apply. Default is 0. +Allowed range is from -180 to 180. + +@item saturation +Set the saturation shift. Default is 0. +Allowed range is from -1 to 1. + +@item intensity +Set the intensity shift. Default is 0. +Allowed range is from -1 to 1. + +@item colors +Set which primary and complementary colors are going to be adjusted. +This options is set by providing one or multiple values. +This can select multiple colors at once. By default all colors are selected. +@table @samp +@item r +Adjust reds. +@item y +Adjust yellows. +@item g +Adjust greens. +@item c +Adjust cyans. +@item b +Adjust blues. +@item m +Adjust magentas. +@item a +Adjust all colors. +@end table + +@item strength +Set strength of filtering. Allowed range is from 0 to 100. +Default value is 1. + +@item rw, gw, bw +Set weight for each RGB component. Allowed range is from 0 to 1. +By default is set to 0.333, 0.334, 0.333. +Those options are used in saturation and lightess processing. + +@item lightness +Set preserving lightness, by default is disabled. +Adjusting hues can change lightness from original RGB triplet, +with this option enabled lightness is kept at same value. +@end table + @section hysteresis Grow first stream into second stream by connecting components. diff --git a/libavfilter/Makefile b/libavfilter/Makefile index e24f419834..933d9628ea 100644 --- a/libavfilter/Makefile +++ b/libavfilter/Makefile @@ -305,6 +305,7 @@ OBJS-$(CONFIG_HSTACK_FILTER) += vf_stack.o framesync.o OBJS-$(CONFIG_HSVHOLD_FILTER) += vf_hsvkey.o OBJS-$(CONFIG_HSVKEY_FILTER) += vf_hsvkey.o OBJS-$(CONFIG_HUE_FILTER) += vf_hue.o +OBJS-$(CONFIG_HUESATURATION_FILTER) += vf_huesaturation.o OBJS-$(CONFIG_HWDOWNLOAD_FILTER) += vf_hwdownload.o OBJS-$(CONFIG_HWMAP_FILTER) += vf_hwmap.o OBJS-$(CONFIG_HWUPLOAD_CUDA_FILTER) += vf_hwupload_cuda.o diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c index 9c99f816d3..409ab5d3c4 100644 --- a/libavfilter/allfilters.c +++ b/libavfilter/allfilters.c @@ -290,6 +290,7 @@ extern const AVFilter ff_vf_hstack; extern const AVFilter ff_vf_hsvhold; extern const AVFilter ff_vf_hsvkey; extern const AVFilter ff_vf_hue; +extern const AVFilter ff_vf_huesaturation; extern const AVFilter ff_vf_hwdownload; extern const AVFilter ff_vf_hwmap; extern const AVFilter ff_vf_hwupload; diff --git a/libavfilter/version.h b/libavfilter/version.h index c04bb7da0d..f6bb4fbe69 100644 --- a/libavfilter/version.h +++ b/libavfilter/version.h @@ -30,7 +30,7 @@ #include "libavutil/version.h" #define LIBAVFILTER_VERSION_MAJOR 8 -#define LIBAVFILTER_VERSION_MINOR 15 +#define LIBAVFILTER_VERSION_MINOR 16 #define LIBAVFILTER_VERSION_MICRO 100 diff --git a/libavfilter/vf_huesaturation.c b/libavfilter/vf_huesaturation.c new file mode 100644 index 0000000000..af7ea33b9e --- /dev/null +++ b/libavfilter/vf_huesaturation.c @@ -0,0 +1,480 @@ +/* + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/opt.h" +#include "libavutil/imgutils.h" +#include "avfilter.h" +#include "drawutils.h" +#include "formats.h" +#include "internal.h" +#include "video.h" + +#define R 0 +#define G 1 +#define B 2 + +#define REDS 0 +#define YELLOWS 1 +#define GREENS 2 +#define CYANS 3 +#define BLUES 4 +#define MAGENTAS 5 + +#define RED (1 << REDS) +#define YELLOW (1 << YELLOWS) +#define GREEN (1 << GREENS) +#define CYAN (1 << CYANS) +#define BLUE (1 << BLUES) +#define MAGENTA (1 << MAGENTAS) +#define ALL 0x3F + +typedef struct HueSaturationContext { + const AVClass *class; + + float hue; + float saturation; + float intensity; + float strength; + float rlw, glw, blw; + int lightness; + int colors; + + int depth; + int planewidth[4]; + int planeheight[4]; + + float matrix[4][4]; + int64_t imatrix[4][4]; + + int bpp; + int step; + uint8_t rgba_map[4]; + + int (*do_slice[2])(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs); +} HueSaturationContext; + +#define DENOM 0x10000 + +static inline void get_triplet(int64_t m[4][4], int *r, int *g, int *b) +{ + const int ir = *r, ig = *g, ib = *b; + + *r = (ir * m[0][0] + ig * m[1][0] + ib * m[2][0] /*+ m[3][0]*/) >> 16; + *g = (ir * m[0][1] + ig * m[1][1] + ib * m[2][1] /*+ m[3][1]*/) >> 16; + *b = (ir * m[0][2] + ig * m[1][2] + ib * m[2][2] /*+ m[3][2]*/) >> 16; +} + +#define FAST_DIV255(x) ((((x) + 128) * 257) >> 16) + +static inline int lerpi8(int v0, int v1, int f, int max) +{ + return v0 + FAST_DIV255((v1 - v0) * f); +} + +static inline int lerpi16(int v0, int v1, int f, int max) +{ + return v0 + (v1 - v0) * (int64_t)f / max; +} + +#define HUESATURATION(name, type, clip, xall) \ +static int do_slice_##name##_##xall(AVFilterContext *ctx, \ + void *arg, \ + int jobnr, int nb_jobs) \ +{ \ + HueSaturationContext *s = ctx->priv; \ + AVFrame *frame = arg; \ + const int imax = (1 << name) - 1; \ + const float strength = s->strength; \ + const int colors = s->colors; \ + const int step = s->step; \ + const int width = frame->width; \ + const int process_h = frame->height; \ + const int slice_start = (process_h * jobnr ) / nb_jobs; \ + const int slice_end = (process_h * (jobnr+1)) / nb_jobs; \ + const int linesize = frame->linesize[0] / sizeof(type); \ + type *row = (type *)frame->data[0] + linesize * slice_start; \ + const uint8_t offset_r = s->rgba_map[R]; \ + const uint8_t offset_g = s->rgba_map[G]; \ + const uint8_t offset_b = s->rgba_map[B]; \ + type *dst_r = row + offset_r; \ + type *dst_g = row + offset_g; \ + type *dst_b = row + offset_b; \ + \ + for (int y = slice_start; y < slice_end; y++) { \ + for (int x = 0; x < width * step; x += step) { \ + int ir, ig, ib, ro, go, bo; \ + \ + ir = ro = dst_r[x]; \ + ig = go = dst_g[x]; \ + ib = bo = dst_b[x]; \ + \ + if (xall) { \ + get_triplet(s->imatrix, &ir, &ig, &ib); \ + } else { \ + const int min = FFMIN3(ir, ig, ib); \ + const int max = FFMAX3(ir, ig, ib); \ + const int flags = (ir == max) << REDS \ + | (ir == min) << CYANS \ + | (ig == max) << GREENS \ + | (ig == min) << MAGENTAS \ + | (ib == max) << BLUES \ + | (ib == min) << YELLOWS; \ + if (colors & flags) { \ + int f = 0; \ + \ + if (colors & RED) \ + f = FFMAX(f, ir - FFMAX(ig, ib)); \ + if (colors & YELLOW) \ + f = FFMAX(f, FFMIN(ir, ig) - ib); \ + if (colors & GREEN) \ + f = FFMAX(f, ig - FFMAX(ir, ib)); \ + if (colors & CYAN) \ + f = FFMAX(f, FFMIN(ig, ib) - ir); \ + if (colors & BLUE) \ + f = FFMAX(f, ib - FFMAX(ir, ig)); \ + if (colors & MAGENTA) \ + f = FFMAX(f, FFMIN(ir, ib) - ig); \ + f = FFMIN(f * strength, imax); \ + get_triplet(s->imatrix, &ir, &ig, &ib); \ + ir = lerpi##name(ro, ir, f, imax); \ + ig = lerpi##name(go, ig, f, imax); \ + ib = lerpi##name(bo, ib, f, imax); \ + } \ + } \ + \ + dst_r[x] = clip(ir); \ + dst_g[x] = clip(ig); \ + dst_b[x] = clip(ib); \ + } \ + \ + dst_r += linesize; \ + dst_g += linesize; \ + dst_b += linesize; \ + } \ + \ + return 0; \ +} + +HUESATURATION(8, uint8_t, av_clip_uint8, 0) +HUESATURATION(16, uint16_t, av_clip_uint16, 0) + +HUESATURATION(8, uint8_t, av_clip_uint8, 1) +HUESATURATION(16, uint16_t, av_clip_uint16, 1) + +static void identity_matrix(float matrix[4][4]) +{ + for (int y = 0; y < 4; y++) + for (int x = 0; x < 4; x++) + matrix[y][x] = y == x; +} + +static void matrix_multiply(float a[4][4], float b[4][4], float c[4][4]) +{ + float temp[4][4]; + + for (int y = 0; y < 4; y++) { + for (int x = 0; x < 4; x++) { + temp[y][x] = b[y][0] * a[0][x] + + b[y][1] * a[1][x] + + b[y][2] * a[2][x] + + b[y][3] * a[3][x]; + } + } + + for (int y = 0; y < 4; y++) { + for (int x = 0; x < 4; x++) + c[y][x] = temp[y][x]; + } +} + +static void colorscale_matrix(float matrix[4][4], float r, float g, float b) +{ + float temp[4][4]; + + temp[0][0] = r; temp[0][1] = 0.f; temp[0][2] = 0.f; temp[0][3] = 0.f; + temp[1][0] = 0.f; temp[1][1] = g; temp[1][2] = 0.f; temp[1][3] = 0.f; + temp[2][0] = 0.f; temp[2][1] = 0.f; temp[2][2] = b; temp[2][3] = 0.f; + temp[3][0] = 0.f; temp[3][1] = 0.f; temp[3][2] = 0.f; temp[3][3] = 1.f; + + matrix_multiply(temp, matrix, matrix); +} + +static void saturation_matrix(float matrix[4][4], float saturation, + float rlw, float glw, float blw) +{ + float s = 1.f - saturation; + float a = s * rlw + saturation; + float b = s * rlw; + float c = s * rlw; + float d = s * glw; + float e = s * glw + saturation; + float f = s * glw; + float g = s * blw; + float h = s * blw; + float i = s * blw + saturation; + float m[4][4]; + + m[0][0] = a; m[0][1] = b; m[0][2] = c; m[0][3] = 0.f; + m[1][0] = d; m[1][1] = e; m[1][2] = f; m[1][3] = 0.f; + m[2][0] = g; m[2][1] = h; m[2][2] = i; m[2][3] = 0.f; + m[3][0] = 0.f; m[3][1] = 0.f; m[3][2] = 0.f; m[3][3] = 1.f; + + matrix_multiply(m, matrix, matrix); +} + +static void matrix2imatrix(float matrix[4][4], int64_t imatrix[4][4]) +{ + for (int y = 0; y < 4; y++) + for (int x = 0; x < 4; x++) + imatrix[y][x] = lrintf(matrix[y][x] * DENOM); +} + +static void x_rotate_matrix(float matrix[4][4], float rs, float rc) +{ + float m[4][4]; + + m[0][0] = 1.f; m[0][1] = 0.f; m[0][2] = 0.f; m[0][3] = 0.f; + m[1][0] = 0.f; m[1][1] = rc; m[1][2] = rs; m[1][3] = 0.f; + m[2][0] = 0.f; m[2][1] = -rs; m[2][2] = rc; m[2][3] = 0.f; + m[3][0] = 0.f; m[3][1] = 0.f; m[3][2] = 0.f; m[3][3] = 1.f; + + matrix_multiply(m, matrix, matrix); +} + +static void y_rotate_matrix(float matrix[4][4], float rs, float rc) +{ + float m[4][4]; + + m[0][0] = rc; m[0][1] = 0.f; m[0][2] = -rs; m[0][3] = 0.f; + m[1][0] = 0.f; m[1][1] = 1.f; m[1][2] = 0.f; m[1][3] = 0.f; + m[2][0] = rs; m[2][1] = 0.f; m[2][2] = rc; m[2][3] = 0.f; + m[3][0] = 0.f; m[3][1] = 0.f; m[3][2] = 0.f; m[3][3] = 1.f; + + matrix_multiply(m, matrix, matrix); +} + +static void z_rotate_matrix(float matrix[4][4], float rs, float rc) +{ + float m[4][4]; + + m[0][0] = rc; m[0][1] = rs; m[0][2] = 0.f; m[0][3] = 0.f; + m[1][0] = -rs; m[1][1] = rc; m[1][2] = 0.f; m[1][3] = 0.f; + m[2][0] = 0.f; m[2][1] = 0.f; m[2][2] = 1.f; m[2][3] = 0.f; + m[3][0] = 0.f; m[3][1] = 0.f; m[3][2] = 0.f; m[3][3] = 1.f; + + matrix_multiply(m, matrix, matrix); +} + +static void z_shear_matrix(float matrix[4][4], float dx, float dy) +{ + float m[4][4]; + + m[0][0] = 1.f; m[0][1] = 0.f; m[0][2] = dx; m[0][3] = 0.f; + m[1][0] = 0.f; m[1][1] = 1.f; m[1][2] = dy; m[1][3] = 0.f; + m[2][0] = 0.f; m[2][1] = 0.f; m[2][2] = 1.f; m[2][3] = 0.f; + m[3][0] = 0.f; m[3][1] = 0.f; m[3][2] = 0.f; m[3][3] = 1.f; + + matrix_multiply(m, matrix, matrix); +} + +static void transform_point(float matrix[4][4], + float x, float y, float z, + float *tx, float *ty, float *tz) +{ + x = y; + *tx = x; + *tx = x * matrix[0][0] + y * matrix[1][0] + z * matrix[2][0] + matrix[3][0]; + *ty = x * matrix[0][1] + y * matrix[1][1] + z * matrix[2][1] + matrix[3][1]; + *tz = x * matrix[0][2] + y * matrix[1][2] + z * matrix[2][2] + matrix[3][2]; +} + +static void hue_rotate_matrix(float matrix[4][4], float rotation, + float rlw, float glw, float blw) +{ + float mag, lx, ly, lz; + float xrs, xrc; + float yrs, yrc; + float zrs, zrc; + float zsx, zsy; + + mag = M_SQRT2; + xrs = 1.f / mag; + xrc = 1.f / mag; + x_rotate_matrix(matrix, xrs, xrc); + + mag = sqrtf(3.f); + yrs = -1.f / mag; + yrc = M_SQRT2 / mag; + y_rotate_matrix(matrix, yrs, yrc); + + transform_point(matrix, rlw, glw, blw, &lx, &ly, &lz); + zsx = lx / lz; + zsy = ly / lz; + z_shear_matrix(matrix, zsx, zsy); + + zrs = sinf(rotation * M_PI / 180.f); + zrc = cosf(rotation * M_PI / 180.f); + z_rotate_matrix(matrix, zrs, zrc); + + z_shear_matrix(matrix, -zsx, -zsy); + + y_rotate_matrix(matrix, -yrs, yrc); + x_rotate_matrix(matrix, -xrs, xrc); +} + +static void shue_rotate_matrix(float m[4][4], float rotation) +{ + float xrs, xrc, yrs, yrc, zrs, zrc, mag; + + mag = M_SQRT2; + xrs = 1.f / mag; + xrc = 1.f / mag; + x_rotate_matrix(m, xrs, xrc); + + mag = sqrtf(3.f); + yrs = -1.f / mag; + yrc = M_SQRT2 / mag; + y_rotate_matrix(m, yrs, yrc); + + zrs = sinf(rotation * M_PI / 180.f); + zrc = cosf(rotation * M_PI / 180.f); + z_rotate_matrix(m, zrs, zrc); + + y_rotate_matrix(m, -yrs, yrc); + x_rotate_matrix(m, -xrs, xrc); +} + +static void init_matrix(HueSaturationContext *s) +{ + float i = 1.f + s->intensity; + float saturation = 1.f + s->saturation; + float hue = s->hue; + + identity_matrix(s->matrix); + colorscale_matrix(s->matrix, i, i, i); + saturation_matrix(s->matrix, saturation, + s->rlw, s->glw, s->blw); + + if (s->lightness) + hue_rotate_matrix(s->matrix, hue, + s->rlw, s->glw, s->blw); + else + shue_rotate_matrix(s->matrix, hue); + + matrix2imatrix(s->matrix, s->imatrix); +} + +static int filter_frame(AVFilterLink *inlink, AVFrame *frame) +{ + AVFilterContext *ctx = inlink->dst; + HueSaturationContext *s = ctx->priv; + + init_matrix(s); + + ff_filter_execute(ctx, s->do_slice[(s->strength >= 99.f) && (s->colors == ALL)], frame, NULL, + FFMIN(s->planeheight[1], ff_filter_get_nb_threads(ctx))); + + return ff_filter_frame(ctx->outputs[0], frame); +} + +static const enum AVPixelFormat pixel_fmts[] = { + AV_PIX_FMT_RGB24, AV_PIX_FMT_BGR24, + AV_PIX_FMT_RGBA, AV_PIX_FMT_BGRA, + AV_PIX_FMT_ABGR, AV_PIX_FMT_ARGB, + AV_PIX_FMT_0BGR, AV_PIX_FMT_0RGB, + AV_PIX_FMT_RGB0, AV_PIX_FMT_BGR0, + AV_PIX_FMT_RGB48, AV_PIX_FMT_BGR48, + AV_PIX_FMT_RGBA64, AV_PIX_FMT_BGRA64, + AV_PIX_FMT_NONE +}; + +static av_cold int config_input(AVFilterLink *inlink) +{ + AVFilterContext *ctx = inlink->dst; + HueSaturationContext *s = ctx->priv; + const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format); + + s->depth = desc->comp[0].depth; + s->bpp = s->depth >> 3; + s->step = av_get_padded_bits_per_pixel(desc) >> (3 + (s->bpp == 2)); + ff_fill_rgba_map(s->rgba_map, inlink->format); + + s->planewidth[1] = s->planewidth[2] = AV_CEIL_RSHIFT(inlink->w, desc->log2_chroma_w); + s->planewidth[0] = s->planewidth[3] = inlink->w; + s->planeheight[1] = s->planeheight[2] = AV_CEIL_RSHIFT(inlink->h, desc->log2_chroma_h); + s->planeheight[0] = s->planeheight[3] = inlink->h; + + s->do_slice[0] = s->depth <= 8 ? do_slice_8_0 : do_slice_16_0; + s->do_slice[1] = s->depth <= 8 ? do_slice_8_1 : do_slice_16_1; + + return 0; +} + +static const AVFilterPad huesaturation_inputs[] = { + { + .name = "default", + .type = AVMEDIA_TYPE_VIDEO, + .flags = AVFILTERPAD_FLAG_NEEDS_WRITABLE, + .filter_frame = filter_frame, + .config_props = config_input, + }, +}; + +static const AVFilterPad huesaturation_outputs[] = { + { + .name = "default", + .type = AVMEDIA_TYPE_VIDEO, + }, +}; + +#define OFFSET(x) offsetof(HueSaturationContext, x) +#define VF AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_RUNTIME_PARAM + +static const AVOption huesaturation_options[] = { + { "hue", "set the hue shift", OFFSET(hue), AV_OPT_TYPE_FLOAT, {.dbl=0},-180, 180, VF }, + { "saturation", "set the saturation shift", OFFSET(saturation), AV_OPT_TYPE_FLOAT, {.dbl=0}, -1, 1, VF }, + { "intensity", "set the intensity shift", OFFSET(intensity), AV_OPT_TYPE_FLOAT, {.dbl=0}, -1, 1, VF }, + { "colors", "set colors range", OFFSET(colors), AV_OPT_TYPE_FLAGS, {.i64=ALL}, 0,ALL,VF, "colors" }, + { "r", "set reds", 0, AV_OPT_TYPE_CONST, {.i64=RED}, 0, 0, VF, "colors" }, + { "y", "set yellows", 0, AV_OPT_TYPE_CONST, {.i64=YELLOW}, 0, 0, VF, "colors" }, + { "g", "set greens", 0, AV_OPT_TYPE_CONST, {.i64=GREEN}, 0, 0, VF, "colors" }, + { "c", "set cyans", 0, AV_OPT_TYPE_CONST, {.i64=CYAN}, 0, 0, VF, "colors" }, + { "b", "set blues", 0, AV_OPT_TYPE_CONST, {.i64=BLUE}, 0, 0, VF, "colors" }, + { "m", "set magentas", 0, AV_OPT_TYPE_CONST, {.i64=MAGENTA}, 0, 0, VF, "colors" }, + { "a", "set all colors", 0, AV_OPT_TYPE_CONST, {.i64=ALL}, 0, 0, VF, "colors" }, + { "strength", "set the filtering strength", OFFSET(strength), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0,100,VF }, + { "rw", "set the red weight", OFFSET(rlw), AV_OPT_TYPE_FLOAT, {.dbl=.333}, 0, 1, VF }, + { "gw", "set the green weight", OFFSET(glw), AV_OPT_TYPE_FLOAT, {.dbl=.334}, 0, 1, VF }, + { "bw", "set the blue weight", OFFSET(blw), AV_OPT_TYPE_FLOAT, {.dbl=.333}, 0, 1, VF }, + { "lightness", "set the preserve lightness", OFFSET(lightness), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, VF }, + { NULL } +}; + +AVFILTER_DEFINE_CLASS(huesaturation); + +const AVFilter ff_vf_huesaturation = { + .name = "huesaturation", + .description = NULL_IF_CONFIG_SMALL("Apply hue-saturation-intensity adjustments."), + .priv_size = sizeof(HueSaturationContext), + .priv_class = &huesaturation_class, + FILTER_INPUTS(huesaturation_inputs), + FILTER_OUTPUTS(huesaturation_outputs), + FILTER_PIXFMTS_ARRAY(pixel_fmts), + .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC | AVFILTER_FLAG_SLICE_THREADS, + .process_command = ff_filter_process_command, +};

1 0

configure: add initial RISC-V support
by Brad Smith 29 Oct '21

29 Oct '21

ffmpeg | branch: master | Brad Smith <brad(a)comstyle.com> | Sun Jul 25 18:31:27 2021 -0400| [dbfafe11ae7239e242f9f328fd1e6bdc620e1dba] | committer: Michael Niedermayer configure: add initial RISC-V support OpenBSD only supports riscv64 but this is an attempt at adding some of the initial bits for RISC-V support. Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=dbfafe11ae7239e24… --- configure | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/configure b/configure index 0fe39375ad..ede8f9777b 100755 --- a/configure +++ b/configure @@ -2026,6 +2026,7 @@ ARCH_LIST=" parisc ppc ppc64 + riscv s390 sh4 sparc @@ -2636,7 +2637,7 @@ for ext in $(filter_out mmx $ARCH_EXT_LIST_X86_SIMD); do done aligned_stack_if_any="aarch64 ppc x86" -fast_64bit_if_any="aarch64 alpha ia64 mips64 parisc64 ppc64 sparc64 x86_64" +fast_64bit_if_any="aarch64 alpha ia64 mips64 parisc64 ppc64 riscv64 sparc64 x86_64" fast_clz_if_any="aarch64 alpha avr32 mips ppc x86" fast_unaligned_if_any="aarch64 ppc x86" simd_align_16_if_any="altivec neon sse" @@ -4945,6 +4946,9 @@ case "$arch" in "Power Macintosh"|ppc*|powerpc*) arch="ppc" ;; + riscv*) + arch="riscv" + ;; s390|s390x) arch="s390" ;; @@ -5338,6 +5342,10 @@ case "$arch" in check_64bit ppc ppc64 enabled shared && enable_weak pic ;; + riscv) + check_64bit riscv32 riscv64 + enabled shared && enable_weak pic + ;; s390) check_64bit s390 s390x enabled shared && enable_weak pic

1 0

fate: update ref files after a recent change
by James Almer 29 Oct '21

29 Oct '21

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Fri Oct 29 15:43:14 2021 -0300| [ba0ba5539af7ffec19a060e6c1073ff106d77c08] | committer: James Almer fate: update ref files after a recent change Missed in 3925b826df365a2365118600bd7910cbca95286d. Signed-off-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=ba0ba5539af7ffec1… --- tests/ref/fate/iv8-demux | 2 +- tests/ref/fate/segment-mp4-to-ts | 8 ++++---- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/tests/ref/fate/iv8-demux b/tests/ref/fate/iv8-demux index 5b4cf0d4b0..1ca76a49c3 100644 --- a/tests/ref/fate/iv8-demux +++ b/tests/ref/fate/iv8-demux @@ -27,5 +27,5 @@ 0, 72000, 72000, 3600, 20891, 0x3d064fd3 0, 75600, 75600, 3600, 20834, 0xcb774dbc 0, 79200, 79200, 3600, 20870, 0xbc536589 -0, 82800, 82800, 3600, 21421, 0xc99a68e4 +0, 82800, 82800, 3600, 21421, 0xc99a68e4, F=0x3 0, 86400, 86400, 3600, 12869, 0x5684e304 diff --git a/tests/ref/fate/segment-mp4-to-ts b/tests/ref/fate/segment-mp4-to-ts index 8b0746fa92..54fcadb846 100644 --- a/tests/ref/fate/segment-mp4-to-ts +++ b/tests/ref/fate/segment-mp4-to-ts @@ -23,7 +23,7 @@ 0, 50400, 54000, 3600, 607, 0xc53c2339, F=0x0, S=1, 1 0, 54000, 72000, 3600, 4755, 0x2f642b58, F=0x0, S=1, 1 0, 57600, 64800, 3600, 1182, 0xbe1a4847, F=0x0, S=1, 1 -0, 61200, 61200, 3600, 809, 0x8d948a4e, F=0x0, S=1, 1 +0, 61200, 61200, 3600, 809, 0x8d948a4e, F=0x2, S=1, 1 0, 64800, 68400, 3600, 656, 0x4fa03c2b, F=0x0, S=1, 1 0, 68400, 86400, 3600, 26555, 0x5629b584, S=1, 1 0, 72000, 79200, 3600, 1141, 0x761b31e8, F=0x0, S=1, 1 @@ -47,7 +47,7 @@ 0, 136800, 140400, 3600, 754, 0xab1c815e, F=0x0, S=1, 1 0, 140400, 158400, 3600, 5328, 0xd2c55ac6, F=0x0, S=1, 1 0, 144000, 151200, 3600, 1271, 0x46006870, F=0x0, S=1, 1 -0, 147600, 147600, 3600, 849, 0x94dc99c7, F=0x0, S=1, 1 +0, 147600, 147600, 3600, 849, 0x94dc99c7, F=0x2, S=1, 1 0, 151200, 154800, 3600, 753, 0xf4236cab, F=0x0, S=1, 1 0, 154800, 172800, 3600, 25825, 0xd5464dee, S=1, 1 0, 158400, 165600, 3600, 1206, 0x8ce84344, F=0x0, S=1, 1 @@ -95,7 +95,7 @@ 0, 309600, 313200, 3600, 829, 0xffd795cd, F=0x0, S=1, 1 0, 313200, 331200, 3600, 5352, 0x59997996, F=0x0, S=1, 1 0, 316800, 324000, 3600, 1501, 0xb3b8f001, F=0x0, S=1, 1 -0, 320400, 320400, 3600, 941, 0x92b0cb18, F=0x0, S=1, 1 +0, 320400, 320400, 3600, 941, 0x92b0cb18, F=0x2, S=1, 1 0, 324000, 327600, 3600, 823, 0x3d548355, F=0x0, S=1, 1 0, 327600, 345600, 3600, 24042, 0x441e94fb, S=1, 1 0, 331200, 338400, 3600, 1582, 0x4f5d1049, F=0x0, S=1, 1 @@ -119,7 +119,7 @@ 0, 396000, 399600, 3600, 199, 0x79b06355, F=0x0, S=1, 1 0, 399600, 417600, 3600, 1862, 0x22a2a06c, F=0x0, S=1, 1 0, 403200, 410400, 3600, 359, 0x11bdae52, F=0x0, S=1, 1 -0, 406800, 406800, 3600, 235, 0xbec26964, F=0x0, S=1, 1 +0, 406800, 406800, 3600, 235, 0xbec26964, F=0x2, S=1, 1 0, 410400, 414000, 3600, 221, 0x8380682c, F=0x0, S=1, 1 0, 414000, 432000, 3600, 22588, 0xf0ecf072, S=1, 1 0, 417600, 424800, 3600, 383, 0x4f3bb571, F=0x0, S=1, 1

1 0

avfilter/vf_nlmeans: split wa struct
by Paul B Mahol 29 Oct '21

29 Oct '21

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Fri Oct 29 16:56:35 2021 +0200| [2b6d290fbf1745c3a562247d8314dd9bab1b9c18] | committer: Paul B Mahol avfilter/vf_nlmeans: split wa struct This will make x86 SIMD simpler and faster. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=2b6d290fbf1745c3a… --- libavfilter/vf_nlmeans.c | 43 +++++++++++++++++++++++++------------------ libavfilter/vf_nlmeans.h | 8 ++------ 2 files changed, 27 insertions(+), 24 deletions(-) diff --git a/libavfilter/vf_nlmeans.c b/libavfilter/vf_nlmeans.c index 93a14bcf19..dee1f68101 100644 --- a/libavfilter/vf_nlmeans.c +++ b/libavfilter/vf_nlmeans.c @@ -52,8 +52,9 @@ typedef struct NLMeansContext { uint32_t *ii; // integral image starting after the 0-line and 0-column int ii_w, ii_h; // width and height of the integral image ptrdiff_t ii_lz_32; // linesize in 32-bit units of the integral image - struct weighted_avg *wa; // weighted average of every pixel - ptrdiff_t wa_linesize; // linesize for wa in struct size unit + float *total_weight; // total weight for every pixel + float *sum; // weighted sum for every pixel + int linesize; // sum and total_weight linesize float *weight_lut; // lookup table mapping (scaled) patch differences to their associated weights uint32_t max_meaningful_diff; // maximum difference considered (if the patch difference is too high we ignore the pixel) NLMeansDSPContext dsp; @@ -307,9 +308,10 @@ static int config_input(AVFilterLink *inlink) s->ii = s->ii_orig + s->ii_lz_32 + 1; // allocate weighted average for every pixel - s->wa_linesize = inlink->w; - s->wa = av_malloc_array(s->wa_linesize, inlink->h * sizeof(*s->wa)); - if (!s->wa) + s->linesize = inlink->w; + s->total_weight = av_malloc_array(inlink->w, inlink->h * sizeof(*s->total_weight)); + s->sum = av_malloc_array(inlink->w, inlink->h * sizeof(*s->sum)); + if (!s->total_weight || !s->sum) return AVERROR(ENOMEM); return 0; @@ -329,7 +331,8 @@ static void compute_weights_line_c(const uint32_t *const iia, const uint32_t *const iid, const uint32_t *const iie, const uint8_t *const src, - struct weighted_avg *wa, + float *total_weight, + float *sum, const float *const weight_lut, int max_meaningful_diff, int startx, int endx) @@ -371,8 +374,8 @@ static void compute_weights_line_c(const uint32_t *const iia, const uint32_t patch_diff_sq = FFMIN(e - d - b + a, max_meaningful_diff); const float weight = weight_lut[patch_diff_sq]; // exp(-patch_diff_sq * s->pdiff_scale) - wa[x].total_weight += weight; - wa[x].sum += weight * src[x]; + total_weight[x] += weight; + sum[x] += weight * src[x]; } } @@ -397,13 +400,14 @@ static int nlmeans_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs for (int y = starty; y < endy; y++) { const uint8_t *const src = td->src + y*src_linesize; - struct weighted_avg *wa = s->wa + y*s->wa_linesize; + float *total_weight = s->total_weight + y*s->linesize; + float *sum = s->sum + y*s->linesize; const uint32_t *const iia = ii; const uint32_t *const iib = ii + dist_b; const uint32_t *const iid = ii + dist_d; const uint32_t *const iie = ii + dist_e; - dsp->compute_weights_line(iia, iib, iid, iie, src, wa, + dsp->compute_weights_line(iia, iib, iid, iie, src, total_weight, sum, weight_lut, max_meaningful_diff, td->startx, td->endx); ii += s->ii_lz_32; @@ -413,19 +417,20 @@ static int nlmeans_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs static void weight_averages(uint8_t *dst, ptrdiff_t dst_linesize, const uint8_t *src, ptrdiff_t src_linesize, - struct weighted_avg *wa, ptrdiff_t wa_linesize, + float *total_weight, float *sum, ptrdiff_t linesize, int w, int h) { for (int y = 0; y < h; y++) { for (int x = 0; x < w; x++) { // Also weight the centered pixel - wa[x].total_weight += 1.f; - wa[x].sum += 1.f * src[x]; - dst[x] = av_clip_uint8(wa[x].sum / wa[x].total_weight + 0.5f); + total_weight[x] += 1.f; + sum[x] += 1.f * src[x]; + dst[x] = av_clip_uint8(sum[x] / total_weight[x] + 0.5f); } dst += dst_linesize; src += src_linesize; - wa += wa_linesize; + total_weight += linesize; + sum += linesize; } } @@ -440,7 +445,8 @@ static int nlmeans_plane(AVFilterContext *ctx, int w, int h, int p, int r, /* focus an integral pointer on the centered image (s1) */ const uint32_t *centered_ii = s->ii + e*s->ii_lz_32 + e; - memset(s->wa, 0, s->wa_linesize * h * sizeof(*s->wa)); + memset(s->total_weight, 0, s->linesize * h * sizeof(*s->total_weight)); + memset(s->sum, 0, s->linesize * h * sizeof(*s->sum)); for (int offy = -r; offy <= r; offy++) { for (int offx = -r; offx <= r; offx++) { @@ -466,7 +472,7 @@ static int nlmeans_plane(AVFilterContext *ctx, int w, int h, int p, int r, } weight_averages(dst, dst_linesize, src, src_linesize, - s->wa, s->wa_linesize, w, h); + s->total_weight, s->sum, s->linesize, w, h); return 0; } @@ -556,7 +562,8 @@ static av_cold void uninit(AVFilterContext *ctx) NLMeansContext *s = ctx->priv; av_freep(&s->weight_lut); av_freep(&s->ii_orig); - av_freep(&s->wa); + av_freep(&s->total_weight); + av_freep(&s->sum); } static const AVFilterPad nlmeans_inputs[] = { diff --git a/libavfilter/vf_nlmeans.h b/libavfilter/vf_nlmeans.h index d0d0056163..cd1ee7c0bf 100644 --- a/libavfilter/vf_nlmeans.h +++ b/libavfilter/vf_nlmeans.h @@ -22,11 +22,6 @@ #include <stddef.h> #include <stdint.h> -struct weighted_avg { - float total_weight; - float sum; -}; - typedef struct NLMeansDSPContext { void (*compute_safe_ssd_integral_image)(uint32_t *dst, ptrdiff_t dst_linesize_32, const uint8_t *s1, ptrdiff_t linesize1, @@ -37,7 +32,8 @@ typedef struct NLMeansDSPContext { const uint32_t *const iid, const uint32_t *const iie, const uint8_t *const src, - struct weighted_avg *wa, + float *total_weight, + float *sum, const float *const weight_lut, int max_meaningful_diff, int startx, int endx);

1 0

avfilter/vf_nlmeans: refactor line processing in preparation for x86 SIMD assembly
by Paul B Mahol 29 Oct '21

29 Oct '21

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Sun Oct 24 14:53:28 2021 +0200| [eaca36d495f802ebaa5729f345016463d3071c27] | committer: Paul B Mahol avfilter/vf_nlmeans: refactor line processing in preparation for x86 SIMD assembly > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=eaca36d495f802eba… --- libavfilter/vf_nlmeans.c | 109 +++++++++++++++++++++++++++-------------------- libavfilter/vf_nlmeans.h | 14 ++++++ 2 files changed, 77 insertions(+), 46 deletions(-) diff --git a/libavfilter/vf_nlmeans.c b/libavfilter/vf_nlmeans.c index af165c861c..93a14bcf19 100644 --- a/libavfilter/vf_nlmeans.c +++ b/libavfilter/vf_nlmeans.c @@ -38,11 +38,6 @@ #include "vf_nlmeans.h" #include "video.h" -struct weighted_avg { - float total_weight; - float sum; -}; - typedef struct NLMeansContext { const AVClass *class; int nb_planes; @@ -329,6 +324,58 @@ struct thread_data { int p; }; +static void compute_weights_line_c(const uint32_t *const iia, + const uint32_t *const iib, + const uint32_t *const iid, + const uint32_t *const iie, + const uint8_t *const src, + struct weighted_avg *wa, + const float *const weight_lut, + int max_meaningful_diff, + int startx, int endx) +{ + for (int x = startx; x < endx; x++) { + /* + * M is a discrete map where every entry contains the sum of all the entries + * in the rectangle from the top-left origin of M to its coordinate. In the + * following schema, "i" contains the sum of the whole map: + * + * M = +----------+-----------------+----+ + * | | | | + * | | | | + * | a| b| c| + * +----------+-----------------+----+ + * | | | | + * | | | | + * | | X | | + * | | | | + * | d| e| f| + * +----------+-----------------+----+ + * | | | | + * | g| h| i| + * +----------+-----------------+----+ + * + * The sum of the X box can be calculated with: + * X = e-d-b+a + * + * See https://en.wikipedia.org/wiki/Summed_area_table + * + * The compute*_ssd functions compute the integral image M where every entry + * contains the sum of the squared difference of every corresponding pixels of + * two input planes of the same size as M. + */ + const uint32_t a = iia[x]; + const uint32_t b = iib[x]; + const uint32_t d = iid[x]; + const uint32_t e = iie[x]; + const uint32_t patch_diff_sq = FFMIN(e - d - b + a, max_meaningful_diff); + const float weight = weight_lut[patch_diff_sq]; // exp(-patch_diff_sq * s->pdiff_scale) + + wa[x].total_weight += weight; + wa[x].sum += weight * src[x]; + } +} + static int nlmeans_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) { NLMeansContext *s = ctx->priv; @@ -346,50 +393,19 @@ static int nlmeans_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs const int dist_d = dist_b * s->ii_lz_32; const int dist_e = dist_d + dist_b; const float *const weight_lut = s->weight_lut; + NLMeansDSPContext *dsp = &s->dsp; for (int y = starty; y < endy; y++) { - const uint8_t *src = td->src + y*src_linesize; + const uint8_t *const src = td->src + y*src_linesize; struct weighted_avg *wa = s->wa + y*s->wa_linesize; - for (int x = td->startx; x < td->endx; x++) { - /* - * M is a discrete map where every entry contains the sum of all the entries - * in the rectangle from the top-left origin of M to its coordinate. In the - * following schema, "i" contains the sum of the whole map: - * - * M = +----------+-----------------+----+ - * | | | | - * | | | | - * | a| b| c| - * +----------+-----------------+----+ - * | | | | - * | | | | - * | | X | | - * | | | | - * | d| e| f| - * +----------+-----------------+----+ - * | | | | - * | g| h| i| - * +----------+-----------------+----+ - * - * The sum of the X box can be calculated with: - * X = e-d-b+a - * - * See https://en.wikipedia.org/wiki/Summed_area_table - * - * The compute*_ssd functions compute the integral image M where every entry - * contains the sum of the squared difference of every corresponding pixels of - * two input planes of the same size as M. - */ - const uint32_t a = ii[x]; - const uint32_t b = ii[x + dist_b]; - const uint32_t d = ii[x + dist_d]; - const uint32_t e = ii[x + dist_e]; - const uint32_t patch_diff_sq = FFMIN(e - d - b + a, max_meaningful_diff); - const float weight = weight_lut[patch_diff_sq]; // exp(-patch_diff_sq * s->pdiff_scale) - - wa[x].total_weight += weight; - wa[x].sum += weight * src[x]; - } + const uint32_t *const iia = ii; + const uint32_t *const iib = ii + dist_b; + const uint32_t *const iid = ii + dist_d; + const uint32_t *const iie = ii + dist_e; + + dsp->compute_weights_line(iia, iib, iid, iie, src, wa, + weight_lut, max_meaningful_diff, + td->startx, td->endx); ii += s->ii_lz_32; } return 0; @@ -493,6 +509,7 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *in) void ff_nlmeans_init(NLMeansDSPContext *dsp) { dsp->compute_safe_ssd_integral_image = compute_safe_ssd_integral_image_c; + dsp->compute_weights_line = compute_weights_line_c; if (ARCH_AARCH64) ff_nlmeans_init_aarch64(dsp); diff --git a/libavfilter/vf_nlmeans.h b/libavfilter/vf_nlmeans.h index 0a9aab2928..d0d0056163 100644 --- a/libavfilter/vf_nlmeans.h +++ b/libavfilter/vf_nlmeans.h @@ -22,11 +22,25 @@ #include <stddef.h> #include <stdint.h> +struct weighted_avg { + float total_weight; + float sum; +}; + typedef struct NLMeansDSPContext { void (*compute_safe_ssd_integral_image)(uint32_t *dst, ptrdiff_t dst_linesize_32, const uint8_t *s1, ptrdiff_t linesize1, const uint8_t *s2, ptrdiff_t linesize2, int w, int h); + void (*compute_weights_line)(const uint32_t *const iia, + const uint32_t *const iib, + const uint32_t *const iid, + const uint32_t *const iie, + const uint8_t *const src, + struct weighted_avg *wa, + const float *const weight_lut, + int max_meaningful_diff, + int startx, int endx); } NLMeansDSPContext; void ff_nlmeans_init(NLMeansDSPContext *dsp);

1 0

avfilter/vf_nlmeans: avoid if () to help paralellization
by Paul B Mahol 29 Oct '21

29 Oct '21

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Sun Oct 24 14:40:45 2021 +0200| [38c86e7a02c286becbe847ff7bc13d19363ce15e] | committer: Paul B Mahol avfilter/vf_nlmeans: avoid if () to help paralellization > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=38c86e7a02c286bec… --- libavfilter/vf_nlmeans.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) diff --git a/libavfilter/vf_nlmeans.c b/libavfilter/vf_nlmeans.c index d5a71291af..af165c861c 100644 --- a/libavfilter/vf_nlmeans.c +++ b/libavfilter/vf_nlmeans.c @@ -332,6 +332,7 @@ struct thread_data { static int nlmeans_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) { NLMeansContext *s = ctx->priv; + const uint32_t max_meaningful_diff = s->max_meaningful_diff; const struct thread_data *td = arg; const ptrdiff_t src_linesize = td->src_linesize; const int process_h = td->endy - td->starty; @@ -383,13 +384,11 @@ static int nlmeans_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs const uint32_t b = ii[x + dist_b]; const uint32_t d = ii[x + dist_d]; const uint32_t e = ii[x + dist_e]; - const uint32_t patch_diff_sq = e - d - b + a; + const uint32_t patch_diff_sq = FFMIN(e - d - b + a, max_meaningful_diff); + const float weight = weight_lut[patch_diff_sq]; // exp(-patch_diff_sq * s->pdiff_scale) - if (patch_diff_sq < s->max_meaningful_diff) { - const float weight = weight_lut[patch_diff_sq]; // exp(-patch_diff_sq * s->pdiff_scale) - wa[x].total_weight += weight; - wa[x].sum += weight * src[x]; - } + wa[x].total_weight += weight; + wa[x].sum += weight * src[x]; } ii += s->ii_lz_32; } @@ -506,7 +505,7 @@ static av_cold int init(AVFilterContext *ctx) s->pdiff_scale = 1. / (h * h); s->max_meaningful_diff = log(255.) / s->pdiff_scale; - s->weight_lut = av_calloc(s->max_meaningful_diff, sizeof(*s->weight_lut)); + s->weight_lut = av_calloc(s->max_meaningful_diff + 1, sizeof(*s->weight_lut)); if (!s->weight_lut) return AVERROR(ENOMEM); for (int i = 0; i < s->max_meaningful_diff; i++)

1 0

avfilter/vf_nlmeans: no need to print filter options at info level
by Paul B Mahol 29 Oct '21

29 Oct '21

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Sun Oct 24 14:28:12 2021 +0200| [64d19411eb7fe53eef7ab953d1e5d6f06b04ab45] | committer: Paul B Mahol avfilter/vf_nlmeans: no need to print filter options at info level > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=64d19411eb7fe53ee… --- libavfilter/vf_nlmeans.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavfilter/vf_nlmeans.c b/libavfilter/vf_nlmeans.c index 0962056a6e..d5a71291af 100644 --- a/libavfilter/vf_nlmeans.c +++ b/libavfilter/vf_nlmeans.c @@ -526,7 +526,7 @@ static av_cold int init(AVFilterContext *ctx) s->patch_hsize = s->patch_size / 2; s->patch_hsize_uv = s->patch_size_uv / 2; - av_log(ctx, AV_LOG_INFO, "Research window: %dx%d / %dx%d, patch size: %dx%d / %dx%d\n", + av_log(ctx, AV_LOG_DEBUG, "Research window: %dx%d / %dx%d, patch size: %dx%d / %dx%d\n", s->research_size, s->research_size, s->research_size_uv, s->research_size_uv, s->patch_size, s->patch_size, s->patch_size_uv, s->patch_size_uv);

1 0

avfilter/vf_nlmeans: make access to pointer to lut faster
by Paul B Mahol 29 Oct '21

29 Oct '21

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Sun Oct 24 14:26:35 2021 +0200| [959ec353ce974ff22863927c0a9dd354d0968c85] | committer: Paul B Mahol avfilter/vf_nlmeans: make access to pointer to lut faster > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=959ec353ce974ff22… --- libavfilter/vf_nlmeans.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/libavfilter/vf_nlmeans.c b/libavfilter/vf_nlmeans.c index b8d8bb2ec0..0962056a6e 100644 --- a/libavfilter/vf_nlmeans.c +++ b/libavfilter/vf_nlmeans.c @@ -344,6 +344,7 @@ static int nlmeans_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs const int dist_b = 2*p + 1; const int dist_d = dist_b * s->ii_lz_32; const int dist_e = dist_d + dist_b; + const float *const weight_lut = s->weight_lut; for (int y = starty; y < endy; y++) { const uint8_t *src = td->src + y*src_linesize; @@ -385,7 +386,7 @@ static int nlmeans_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs const uint32_t patch_diff_sq = e - d - b + a; if (patch_diff_sq < s->max_meaningful_diff) { - const float weight = s->weight_lut[patch_diff_sq]; // exp(-patch_diff_sq * s->pdiff_scale) + const float weight = weight_lut[patch_diff_sq]; // exp(-patch_diff_sq * s->pdiff_scale) wa[x].total_weight += weight; wa[x].sum += weight * src[x]; }

1 0