[FFmpeg-devel] [PATCH 1/1] avfilter: Implement gray world color correction Uses log LAB colorspace

Paul B Mahol onemda at gmail.com
Sat Aug 28 20:31:47 EEST 2021


On Fri, Aug 27, 2021 at 11:18 AM Paul Buxton <paulbuxton.mail at googlemail.com>
wrote:

> Signed-off-by: Paul Buxton <paulbuxton.mail at googlemail.com>
> ---
>  doc/filters.texi           |  12 ++
>  libavfilter/Makefile       |   1 +
>  libavfilter/allfilters.c   |   1 +
>  libavfilter/version.h      |   2 +-
>  libavfilter/vf_grayworld.c | 362 +++++++++++++++++++++++++++++++++++++
>  5 files changed, 377 insertions(+), 1 deletion(-)
>  create mode 100644 libavfilter/vf_grayworld.c
>
> diff --git a/doc/filters.texi b/doc/filters.texi
> index 269f2920d1..bb20ff538e 100644
> --- a/doc/filters.texi
> +++ b/doc/filters.texi
> @@ -13054,6 +13054,18 @@ Set upper limit for video rate of output stream,
> Default value is @var{25}.
>  This guarantee that output video frame rate will not be higher than this
> value.
>  @end table
>
> + at section grayworld
> +A color constancy filter that applies color correction based on the
> grayworld assumption
> +
> +See: @url{
> https://www.researchgate.net/publication/275213614_A_New_Color_Correction_Method_for_Underwater_Imaging
> }
> +
> +The algorithm  uses linear light, so input
> +data should be linearized beforehand (and possibly correctly tagged).
> +
> + at example
> +ffmpeg -i INPUT -vf
> zscale=transfer=linear,grayworld,zscale=transfer=bt709,format=yuv420p OUTPUT
> + at end example
> +
>  @section greyedge
>  A color constancy variation filter which estimates scene illumination via
> grey edge algorithm
>  and corrects the scene colors accordingly.
> diff --git a/libavfilter/Makefile b/libavfilter/Makefile
> index 3166e70b58..1d86110d6b 100644
> --- a/libavfilter/Makefile
> +++ b/libavfilter/Makefile
> @@ -287,6 +287,7 @@ OBJS-$(CONFIG_GBLUR_FILTER)                  +=
> vf_gblur.o
>  OBJS-$(CONFIG_GEQ_FILTER)                    += vf_geq.o
>  OBJS-$(CONFIG_GRADFUN_FILTER)                += vf_gradfun.o
>  OBJS-$(CONFIG_GRAPHMONITOR_FILTER)           += f_graphmonitor.o
> +OBJS-$(CONFIG_GRAYWORLD_FILTER)              += vf_grayworld.o
>  OBJS-$(CONFIG_GREYEDGE_FILTER)               += vf_colorconstancy.o
>  OBJS-$(CONFIG_GUIDED_FILTER)                 += vf_guided.o
>  OBJS-$(CONFIG_HALDCLUT_FILTER)               += vf_lut3d.o framesync.o
> diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
> index c49c831203..73e5fe1a41 100644
> --- a/libavfilter/allfilters.c
> +++ b/libavfilter/allfilters.c
> @@ -272,6 +272,7 @@ extern const AVFilter ff_vf_gblur;
>  extern const AVFilter ff_vf_geq;
>  extern const AVFilter ff_vf_gradfun;
>  extern const AVFilter ff_vf_graphmonitor;
> +extern const AVFilter ff_vf_grayworld;
>  extern const AVFilter ff_vf_greyedge;
>  extern const AVFilter ff_vf_guided;
>  extern const AVFilter ff_vf_haldclut;
> diff --git a/libavfilter/version.h b/libavfilter/version.h
> index e9a76c5ac3..e3a86d9b01 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   4
> +#define LIBAVFILTER_VERSION_MINOR   5
>  #define LIBAVFILTER_VERSION_MICRO 100
>
>
> diff --git a/libavfilter/vf_grayworld.c b/libavfilter/vf_grayworld.c
> new file mode 100644
> index 0000000000..b81dc2f28f
> --- /dev/null
> +++ b/libavfilter/vf_grayworld.c
> @@ -0,0 +1,362 @@
> +/*
> + * Copyright (c) 2021 Paul Buxton
> + *
> + * 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
> + */
> +
> +/**
> + * @file
> + * Color correction filter based on
> + *
> https://www.researchgate.net/publication/275213614_A_New_Color_Correction_Method_for_Underwater_Imaging
> + *
> + */
> +
> +#include "libavutil/imgutils.h"
> +#include "libavutil/opt.h"
> +#include "libavutil/pixdesc.h"
> +#include "libavutil/intreadwrite.h"
> +
> +#include "avfilter.h"
> +#include "drawutils.h"
> +#include "formats.h"
> +#include "internal.h"
> +#include "video.h"
> +
> +typedef struct ThreadData {
> +    AVFrame *in, *out;
> +    float l_avg;
> +    float a_avg;
> +    float b_avg;
> +    const AVPixFmtDescriptor *desc;
> +    const AVPixFmtDescriptor *odesc;
> +} ThreadData;
> +
> +typedef struct GrayWorldContext {
> +    const AVClass* class;
> +    float *tmpplab;
> +    int *line_count_pels;
> +    float *line_sum;
> +
> +} GrayWorldContext;
> +
> +#define OFFSET(x) offsetof(GrayWorldContext, x)
> +#define FLAGS AV_OPT_FLAG_FILTERING_PARAM | AV_OPT_FLAG_VIDEO_PARAM |
> AV_OPT_FLAG_RUNTIME_PARAM
> +static const AVOption grayworld_options[] = {NULL};
> +
> +AVFILTER_DEFINE_CLASS(grayworld);
> +
> +void compute_correction(GrayWorldContext *s, ThreadData *td);
> +void apply_matrix(const float matrix[3][3], const float input[3], float
> output[3]);
> +void rgb2lab(const float rgb[3], float lab[3]);
> +void lab2rgb(const float lab[3], float rgb[3]);
> +
> +static int query_formats(AVFilterContext *ctx)
> +{
> +    static const enum AVPixelFormat pix_fmts[] = {
> +        AV_PIX_FMT_GBRPF32,
> +        AV_PIX_FMT_GBRAPF32,
> +        AV_PIX_FMT_NONE,
> +    };
> +    return ff_set_common_formats_from_list(ctx, pix_fmts);
> +}
> +
> +void apply_matrix(const float matrix[3][3], const float input[3], float
> output[3])
> +{
> +    for (int row = 0; row < 3; ++row) {
> +        output[row] = matrix[row][0] * input[0] + matrix[row][1] *
> input[1] + matrix[row][2] * input[2];
> +    }
> +}
> +
> +const float lms2lab[3][3] = {
> +    {0.5774, 0.5774, 0.5774},
> +    {0.40825, 0.40825, -0.816458},
> +    {0.707, -0.707, 0}
> +};
> +
> +const float lab2lms[3][3] = {
> +    {0.57735, 0.40825, 0.707},
> +    {0.57735, 0.40825, -0.707},
> +    {0.57735, -0.8165, 0}
> +};
> +
> +const float rgb2lms[3][3] = {
> +    {0.3811, 0.5783, 0.0402},
> +    {0.1967, 0.7244, 0.0782},
> +    {0.0241, 0.1288, 0.8444}
> +};
> +
> +const float lms2rgb[3][3] = {
> +    {4.4679, -3.5873, 0.1193},
> +    {-1.2186, 2.3809, -0.1624},
> +    {0.0497, -0.2439, 1.2045}
> +};
> +
>

all those tables should be static.


> +/**
> + * Convert from Linear RGB to logspace LAB
> + *
> + * @param rgb Input array of rgb components
> + * @param lab output array of lab components
> + */
> +void rgb2lab(const float rgb[3], float lab[3])
> +{
> +    float lms[3];
> +    apply_matrix(rgb2lms, rgb, lms);
> +    lms[0] = logf(lms[0]);
> +    lms[1] = logf(lms[1]);
> +    lms[2] = logf(lms[2]);
> +    apply_matrix(lms2lab, lms, lab);
> +}
> +
> +/**
> + * Convert from Logspace LAB to Linear RGB
> + *
> + * @param lab input array of lab components
> + * @param rgb output array of rgb components
> + */
> +void lab2rgb(const float lab[3], float rgb[3])
> +{
> +    float lms[3];
> +    apply_matrix(lab2lms, lab, lms);
> +    lms[0] = expf(lms[0]);
> +    lms[1] = expf(lms[1]);
> +    lms[2] = expf(lms[2]);
> +    apply_matrix(lms2rgb, lms, rgb);
> +}
> +
> +/**
> + *
> + * Convert a frame from linear RGB to logspace LAB, and accumulate
> channel totals for each row
> + * Convert from RGB -> lms using equation 4 in color transfer paper.
> + *
> + * @param ctx Filter context
> + * @param arg Thread data pointer
> + * @param jobnr job number
> + * @param nb_jobs number of jobs
> + */
> +static int convert_frame(AVFilterContext *ctx, void *arg, int jobnr, int
> nb_jobs)
> +{
> +    GrayWorldContext *s = ctx->priv;
> +    ThreadData *td = arg;
> +    AVFrame *in = td->in;
> +    AVFrame *out = td->out;
> +    AVFilterLink *outlink = ctx->outputs[0];
> +    const int slice_start = (out->height * jobnr) / nb_jobs;
> +    const int slice_end = (out->height * (jobnr + 1)) / nb_jobs;
> +    float rgb[3];
> +    float lab[3];
> +
> +    for (unsigned int i = slice_start; i < slice_end; i++) {
> +        float *b_in_row = (float*)(in->data[1] + i * in->linesize[1]);
> +        float *g_in_row = (float*)(in->data[0] + i * in->linesize[0]);
> +        float *r_in_row = (float*)(in->data[2] + i * in->linesize[2]);
> +        float *acur = s->tmpplab + i * outlink->w + outlink->w *
> outlink->h;
> +        float *bcur = s->tmpplab + i * outlink->w + 2 * outlink->w *
> outlink->h;
> +        float *lcur = s->tmpplab + i * outlink->w;
> +
> +        s->line_sum[i] = 0.0f;
> +        s->line_sum[i + outlink->h] = 0.0f;
> +        s->line_count_pels[i] = 0;
> +
> +        for (unsigned int j = 0; j < outlink->w; j++) {
> +            rgb[0] = r_in_row[j];
> +            rgb[1] = g_in_row[j];
> +            rgb[2] = b_in_row[j];
> +            rgb2lab(rgb, lab);
> +            *(lcur++) = lab[0];
> +            *(acur++) = lab[1];
> +            *(bcur++) = lab[2];
> +            s->line_sum[i] += lab[1];
> +            s->line_sum[i + outlink->h] += lab[2];
> +            s->line_count_pels[i]++;
> +        }
> +    }
> +    return 0;
> +}
> +
> +/**
> + *
> + * Sum the channel totals and compute the median for each channel
> + *
> + * @param s Frame context
> + * @param td thread data
> + */
> +void compute_correction(GrayWorldContext *s, ThreadData *td)
> +{
> +    float asum = 0.0f;
> +    float bsum = 0.0f;
> +    float pixels = 0;
> +    for (int y = 0; y < td->out->height; ++y) {
> +        asum += s->line_sum[y];
> +        bsum += s->line_sum[y + td->out->height];
> +        pixels += s->line_count_pels[y];
> +    }
> +    td->a_avg = asum / pixels;
> +    td->b_avg = bsum / pixels;
> +}
> +
> +/**
> + *
> + * Subtract the mean logspace AB values from each pixel.
> + *
> + * @param ctx Filter context
> + * @param arg Thread data pointer
> + * @param jobnr job number
> + * @param nb_jobs number of jobs
> + */
> +static int correct_frame(AVFilterContext *ctx, void *arg, int jobnr, int
> nb_jobs)
> +{
> +    GrayWorldContext *s = ctx->priv;
> +    ThreadData *td = arg;
> +    AVFrame *out = td->out;
> +    AVFilterLink *outlink = ctx->outputs[0];
> +    const int slice_start = (out->height * jobnr) / nb_jobs;
> +    const int slice_end = (out->height * (jobnr + 1)) / nb_jobs;
> +    float rgb[3];
> +    float lab[3];
> +
> +    for (int i = slice_start; i < slice_end; i++) {
> +        float *g_out_row = (float*)(out->data[0] + i * out->linesize[0]);
> +        float *b_out_row = (float*)(out->data[1] + i * out->linesize[1]);
> +        float *r_out_row = (float*)(out->data[2] + i * out->linesize[2]);
> +        float *lcur = s->tmpplab + i * outlink->w;
> +        float *acur = s->tmpplab + i * outlink->w + outlink->w *
> outlink->h;
> +        float *bcur = s->tmpplab + i * outlink->w + 2 * outlink->w *
> outlink->h;
> +
> +        for (int j = 0; j < outlink->w; j++) {
> +            lab[0] = *lcur++;
> +            lab[1] = *acur++;
> +            lab[2] = *bcur++;
> +
> +            // subtract the average for the color channels
> +            lab[1] -= td->a_avg;
> +            lab[2] -= td->b_avg;
> +
> +            //convert back to linear rgb
> +            lab2rgb(lab, rgb);
> +            r_out_row[j] = rgb[0];
> +            g_out_row[j] = rgb[1];
> +            b_out_row[j] = rgb[2];
> +        }
> +    }
> +    return 0;
> +}
> +
> +static int config_input(AVFilterLink *inlink)
> +{
> +    GrayWorldContext *s = inlink->dst->priv;
> +
> +    s->tmpplab = av_malloc_array(inlink->h * inlink->w * 3, sizeof(float)
> );
> +    if (!s->tmpplab){
> +        return AVERROR(ENOMEM);
> +    }
> +    FF_ALLOC_TYPED_ARRAY(s->line_count_pels, inlink->h);
> +    if (!s->line_count_pels){
> +        return AVERROR(ENOMEM);
> +    }
> +    FF_ALLOC_TYPED_ARRAY(s->line_sum, inlink->h * 2);
> +    if (!s->line_sum){
> +        return AVERROR(ENOMEM);
> +    }
> +
> +    return 0;
> +}
> +
> +static av_cold void uninit(AVFilterContext *ctx)
> +{
> +    GrayWorldContext *s = ctx->priv;
> +
> +    av_freep(&s->tmpplab);
> +    av_freep(&s->line_count_pels);
> +    av_freep(&s->line_sum);
> +}
> +
> +static int filter_frame(AVFilterLink *inlink, AVFrame *in)
> +{
> +    AVFilterContext *ctx = inlink->dst;
> +    GrayWorldContext *s = ctx->priv;
> +    AVFilterLink *outlink = ctx->outputs[0];
> +    ThreadData td;
> +    AVFrame *out;
> +    const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
> +    const AVPixFmtDescriptor *odesc =
> av_pix_fmt_desc_get(outlink->format);
> +
> +    if (av_frame_is_writable(in)) {
> +        out = in;
> +    } else {
> +        out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
> +        if (!out) {
> +            av_frame_free(&in);
> +            return AVERROR(ENOMEM);
> +        }
> +        av_frame_copy_props(out, in);
> +    }
> +    /* input and output transfer will be linear */
> +    if (in->color_trc == AVCOL_TRC_UNSPECIFIED) {
> +        av_log(s, AV_LOG_WARNING, "Untagged transfer, assuming linear
> light\n");
> +        out->color_trc = AVCOL_TRC_LINEAR;
> +    } else if (in->color_trc != AVCOL_TRC_LINEAR) {
> +        av_log(s, AV_LOG_WARNING, "Grey world color correction works on
> linear light only\n");
> +    }
> +
> +    td.in = in;
> +    td.out = out;
> +    td.desc = desc;
> +    td.odesc = odesc;
> +
> +    ff_filter_execute(ctx, convert_frame, &td, NULL, FFMIN(outlink->h,
> ff_filter_get_nb_threads(ctx)));
> +    compute_correction(s, &td);
> +    ff_filter_execute(ctx, correct_frame, &td, NULL, FFMIN(outlink->h,
> ff_filter_get_nb_threads(ctx)));
> +
> +    if (in != out){
> +        // copy alpha channel from input frame to output
> +        av_image_copy_plane(out->data[3], out->linesize[3],
> +                            in->data[3], in->linesize[3],
> +                            out->linesize[3], outlink->h);
> +        // free the input frame
>

Please remove obvious comments, like this one above.

> +        av_frame_free(&in);
> +    }
> +
> +    return ff_filter_frame(outlink, out);
> +}
> +
> +static const AVFilterPad grayworld_inputs[] = {
> +    {
> +        .name = "default",
> +        .type = AVMEDIA_TYPE_VIDEO,
> +        .filter_frame = filter_frame,
> +        .config_props = config_input,
> +    }
> +};
> +
> +static const AVFilterPad grayworld_outputs[] = {
> +    {
> +        .name = "default",
> +        .type = AVMEDIA_TYPE_VIDEO,
> +    }
> +};
> +
> +const AVFilter ff_vf_grayworld = {
> +    .name = "grayworld",
> +    .description = NULL_IF_CONFIG_SMALL("Adjust white balance using LAB
> gray world algorithm"),
> +    .priv_size = sizeof(GrayWorldContext),
> +    .priv_class = &grayworld_class,
> +    .query_formats = query_formats,
> +    FILTER_INPUTS(grayworld_inputs),
> +    FILTER_OUTPUTS(grayworld_outputs),
> +    .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC |
> AVFILTER_FLAG_SLICE_THREADS,
> +    .uninit = uninit,
>

Why this is not nicely indented like in any other filter?

+};
> --
> 2.17.1
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".
>


More information about the ffmpeg-devel mailing list