From patchwork Wed Mar 3 18:43:17 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Paul B Mahol X-Patchwork-Id: 26078 Return-Path: X-Original-To: patchwork@ffaux-bg.ffmpeg.org Delivered-To: patchwork@ffaux-bg.ffmpeg.org Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org [79.124.17.100]) by ffaux.localdomain (Postfix) with ESMTP id 24B2E44BE73 for ; Wed, 3 Mar 2021 20:43:38 +0200 (EET) Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 0A7C868A8FA; Wed, 3 Mar 2021 20:43:38 +0200 (EET) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from mail-ej1-f48.google.com (mail-ej1-f48.google.com [209.85.218.48]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id 758EE68A700 for ; Wed, 3 Mar 2021 20:43:29 +0200 (EET) Received: by mail-ej1-f48.google.com with SMTP id ci14so25194362ejc.7 for ; Wed, 03 Mar 2021 10:43:29 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=from:to:subject:date:message-id:in-reply-to:references; bh=JGJ9ALH4rUZts6OoYYCIGQpfeoEzoJLXcaAuf9zQroc=; b=fwKvhhhectQv/CDtJ7kzJ3VP42QO1vfw67moDPAXw1pmlRU3i3PM8slp4ygURtbldC SkvH1ALugWOBuL2ATm3Tw/f70ZWYw4yekS5v2sv4DB/u0Vz4AUCKu8qagfdlPfq87Vjy qm2BXyA8JwrHA6SR4okZVjmikTzMWrEr71TMQaymvRMWr30OikfaLDufIuXkO9VABsbi 7jr27ohHX5e5S229RrLQy2dZBEyR5iGXVy7gEhSPQhlF+PacufhK+fpaCxo7UXbq6lAQ S5wy+mFMMFT06Ua8VugrbCPPJOuBpYhl4PI3RTHGtCXdWvZHY3q5jGAW8nQyhgUPSQYt fkhA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:subject:date:message-id:in-reply-to :references; bh=JGJ9ALH4rUZts6OoYYCIGQpfeoEzoJLXcaAuf9zQroc=; b=ja1eo2+cTGpFlrG9LaVGiwk2QqrZjMX36TeKPFgSzxAphf45sKxJ9wd4lFE1Zo3h2R 7hCj78XE4iQgSns5iTdP/tYJtwWMmEuFDvMjBu/qCQacUMoIk30grxmgobWldVwEkivc 1s8HHPgFoFWJbNcNjPpeDKo8BkDYYZ2ljURh8Yrbn7fWjBPwZjCqWYSzIt+s/RWSHE3G p6r8ps4HZoDjsFPI1ijknl8DVcmS1eBbyITtSLUGJSf/68IHzBOsSVrsjiXeC/zqqsh7 d/cjidxjme9PHjwjEG+qum9m8GkEZmYOI52f91WbmhlG7w0+xlb0u+TBORok8kUIb1VU w9wA== X-Gm-Message-State: AOAM533/HMCUO7kq5k/oQwMDFrwfkhS6IWMmEcqL6tCaKftOD3QTnjNB gpQUc0M5U/GgbgfLYAg43ed2FM1TByQ= X-Google-Smtp-Source: ABdhPJwpA+mXSfK7Y/Nue9Rwhm3Xjq3hlWpyeXIZMBqwyfE2isTgjSKCK85CqBxg19F5Bq49oUDobg== X-Received: by 2002:a17:906:380b:: with SMTP id v11mr199241ejc.183.1614797008864; Wed, 03 Mar 2021 10:43:28 -0800 (PST) Received: from localhost.localdomain ([94.250.185.123]) by smtp.gmail.com with ESMTPSA id v13sm24298271ede.5.2021.03.03.10.43.28 for (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Wed, 03 Mar 2021 10:43:28 -0800 (PST) From: Paul B Mahol To: ffmpeg-devel@ffmpeg.org Date: Wed, 3 Mar 2021 19:43:17 +0100 Message-Id: <20210303184317.8809-2-onemda@gmail.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20210303184317.8809-1-onemda@gmail.com> References: <20210303184317.8809-1-onemda@gmail.com> Subject: [FFmpeg-devel] [PATCH 2/2] avfilter: add msad video filter X-BeenThere: ffmpeg-devel@ffmpeg.org X-Mailman-Version: 2.1.20 Precedence: list List-Id: FFmpeg development discussions and patches List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Reply-To: FFmpeg development discussions and patches MIME-Version: 1.0 Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" Signed-off-by: Paul B Mahol --- doc/filters.texi | 21 +++++++ libavfilter/allfilters.c | 1 + libavfilter/vf_identity.c | 115 ++++++++++++++++++++++++++++++++------ 3 files changed, 120 insertions(+), 17 deletions(-) diff --git a/doc/filters.texi b/doc/filters.texi index 38698b6542..eb0a0b6a0b 100644 --- a/doc/filters.texi +++ b/doc/filters.texi @@ -14903,6 +14903,27 @@ Default value for @option{hi} is 64*12, default value for @option{lo} is 64*5, and default value for @option{frac} is 0.33. @end table +@section msad + +Obtain the MSAD (Mean Sum of Absolute Differences) between two input videos. + +This filter takes two input videos. + +Both input videos must have the same resolution and pixel format for +this filter to work correctly. Also it assumes that both inputs +have the same number of frames, which are compared one by one. + +The obtained per component, average, min and max MSAD is printed through +the logging system. + +The filter stores the calculated MSAD of each frame in frame metadata. + +In the below example the input file @file{main.mpg} being processed is compared +with the reference file @file{ref.mpg}. + +@example +ffmpeg -i main.mpg -i ref.mpg -lavfi msad -f null - +@end example @section negate diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c index 4bd8eabd28..0d02c2b748 100644 --- a/libavfilter/allfilters.c +++ b/libavfilter/allfilters.c @@ -320,6 +320,7 @@ extern AVFilter ff_vf_minterpolate; extern AVFilter ff_vf_mix; extern AVFilter ff_vf_monochrome; extern AVFilter ff_vf_mpdecimate; +extern AVFilter ff_vf_msad; extern AVFilter ff_vf_negate; extern AVFilter ff_vf_nlmeans; extern AVFilter ff_vf_nlmeans_opencl; diff --git a/libavfilter/vf_identity.c b/libavfilter/vf_identity.c index 02dda392b3..60d02e328b 100644 --- a/libavfilter/vf_identity.c +++ b/libavfilter/vf_identity.c @@ -32,6 +32,7 @@ #include "framesync.h" #include "internal.h" #include "video.h" +#include "scene_sad.h" typedef struct IdentityContext { const AVClass *class; @@ -39,7 +40,9 @@ typedef struct IdentityContext { double score, min_score, max_score, score_comp[4]; uint64_t nb_frames; int is_rgb; + int is_msad; uint8_t rgba_map[4]; + int max[4]; char comps[4]; int nb_components; int nb_threads; @@ -47,17 +50,14 @@ typedef struct IdentityContext { int planeheight[4]; uint64_t **scores; unsigned (*filter_line)(const uint8_t *buf, const uint8_t *ref, int w); + int (*filter_slice)(AVFilterContext *ctx, void *arg, + int jobnr, int nb_jobs); + ff_scene_sad_fn sad; } IdentityContext; #define OFFSET(x) offsetof(IdentityContext, x) #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM -static const AVOption identity_options[] = { - { NULL } -}; - -FRAMESYNC_DEFINE_CLASS(identity, IdentityContext, fs); - static unsigned identity_line_8bit(const uint8_t *main_line, const uint8_t *ref_line, int outw) { unsigned score = 0; @@ -91,6 +91,34 @@ typedef struct ThreadData { int nb_components; } ThreadData; +static +int compute_images_msad(AVFilterContext *ctx, void *arg, + int jobnr, int nb_jobs) +{ + IdentityContext *s = ctx->priv; + ThreadData *td = arg; + uint64_t *score = td->score[jobnr]; + + for (int c = 0; c < td->nb_components; c++) { + const int outw = td->planewidth[c]; + const int outh = td->planeheight[c]; + const int slice_start = (outh * jobnr) / nb_jobs; + const int slice_end = (outh * (jobnr+1)) / nb_jobs; + const int ref_linesize = td->ref_linesize[c]; + const int main_linesize = td->main_linesize[c]; + const uint8_t *main_line = td->main_data[c] + main_linesize * slice_start; + const uint8_t *ref_line = td->ref_data[c] + ref_linesize * slice_start; + uint64_t m = 0; + + s->sad(main_line, main_linesize, ref_line, ref_linesize, + outw, slice_end - slice_start, &m); + + score[c] = m; + } + + return 0; +} + static int compute_images_identity(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) @@ -121,16 +149,21 @@ int compute_images_identity(AVFilterContext *ctx, void *arg, return 0; } -static void set_meta(AVDictionary **metadata, const char *key, char comp, float d) +static void set_meta(AVFilterContext *ctx, + AVDictionary **metadata, const char *key, char comp, float d) { char value[128]; snprintf(value, sizeof(value), "%f", d); if (comp) { char key2[128]; - snprintf(key2, sizeof(key2), "%s%c", key, comp); + snprintf(key2, sizeof(key2), "lavfi.%s.%s%s%c", + ctx->filter->name, ctx->filter->name, key, comp); av_dict_set(metadata, key2, value, 0); } else { - av_dict_set(metadata, key, value, 0); + char key2[128]; + snprintf(key2, sizeof(key2), "lavfi.%s.%s%s", + ctx->filter->name, ctx->filter->name, key); + av_dict_set(metadata, key2, value, 0); } } @@ -139,7 +172,7 @@ static int do_identity(FFFrameSync *fs) AVFilterContext *ctx = fs->parent; IdentityContext *s = ctx->priv; AVFrame *master, *ref; - double comp_identity[4], score = 0.; + double comp_score[4], score = 0.; uint64_t comp_sum[4] = { 0 }; AVDictionary **metadata; ThreadData td; @@ -163,7 +196,7 @@ static int do_identity(FFFrameSync *fs) td.planeheight[c] = s->planeheight[c]; } - ctx->internal->execute(ctx, compute_images_identity, &td, NULL, FFMIN(s->planeheight[1], s->nb_threads)); + ctx->internal->execute(ctx, s->filter_slice, &td, NULL, FFMIN(s->planeheight[1], s->nb_threads)); for (int j = 0; j < s->nb_threads; j++) { for (int c = 0; c < s->nb_components; c++) @@ -171,10 +204,13 @@ static int do_identity(FFFrameSync *fs) } for (int c = 0; c < s->nb_components; c++) - comp_identity[c] = comp_sum[c] / ((double)s->planewidth[c] * s->planeheight[c]); + comp_score[c] = comp_sum[c] / ((double)s->planewidth[c] * s->planeheight[c]); + + for (int c = 0; c < s->nb_components && s->is_msad; c++) + comp_score[c] /= (double)s->max[c]; for (int c = 0; c < s->nb_components; c++) - score += comp_identity[c]; + score += comp_score[c]; score /= s->nb_components; s->min_score = FFMIN(s->min_score, score); @@ -183,14 +219,14 @@ static int do_identity(FFFrameSync *fs) s->score += score; for (int j = 0; j < s->nb_components; j++) - s->score_comp[j] += comp_identity[j]; + s->score_comp[j] += comp_score[j]; s->nb_frames++; for (int j = 0; j < s->nb_components; j++) { int c = s->is_rgb ? s->rgba_map[j] : j; - set_meta(metadata, "lavfi.identity.identity.", s->comps[j], comp_identity[c]); + set_meta(ctx, metadata, ".", s->comps[j], comp_score[c]); } - set_meta(metadata, "lavfi.identity.identity_avg", 0, score); + set_meta(ctx, metadata, "_avg", 0, score); return ff_filter_frame(ctx->outputs[0], master); } @@ -269,8 +305,19 @@ static int config_input_ref(AVFilterLink *inlink) s->min_score = +INFINITY; s->max_score = -INFINITY; + s->max[0] = (1 << desc->comp[0].depth) - 1; + s->max[1] = (1 << desc->comp[1].depth) - 1; + s->max[2] = (1 << desc->comp[2].depth) - 1; + s->max[3] = (1 << desc->comp[3].depth) - 1; + + s->is_msad = !strcmp(ctx->filter->name, "msad"); + s->filter_slice = !s->is_msad ? compute_images_identity : compute_images_msad; s->filter_line = desc->comp[0].depth > 8 ? identity_line_16bit : identity_line_8bit; + s->sad = ff_scene_sad_get_fn(desc->comp[0].depth <= 8 ? 8 : 16); + if (!s->sad) + return AVERROR(EINVAL); + return 0; } @@ -322,7 +369,8 @@ static av_cold void uninit(AVFilterContext *ctx) av_strlcatf(buf, sizeof(buf), " %c:%f", s->comps[j], s->score_comp[c] / s->nb_frames); } - av_log(ctx, AV_LOG_INFO, "Identity%s average:%f min:%f max:%f\n", + av_log(ctx, AV_LOG_INFO, "%s%s average:%f min:%f max:%f\n", + ctx->filter->name, buf, s->score / s->nb_frames, s->min_score, @@ -356,6 +404,15 @@ static const AVFilterPad identity_outputs[] = { { NULL } }; +static const AVOption options[] = { + { NULL } +}; + +#if CONFIG_IDENTITY_FILTER + +#define identity_options options +FRAMESYNC_DEFINE_CLASS(identity, IdentityContext, fs); + AVFilter ff_vf_identity = { .name = "identity", .description = NULL_IF_CONFIG_SMALL("Calculate the Identity between two video streams."), @@ -370,3 +427,27 @@ AVFilter ff_vf_identity = { .outputs = identity_outputs, .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL | AVFILTER_FLAG_SLICE_THREADS, }; + +#endif /* CONFIG_IDENTITY_FILTER */ + +#if CONFIG_MSAD_FILTER + +#define msad_options options +FRAMESYNC_DEFINE_CLASS(msad, IdentityContext, fs); + +AVFilter ff_vf_msad = { + .name = "msad", + .description = NULL_IF_CONFIG_SMALL("Calculate the MSAD between two video streams."), + .preinit = msad_framesync_preinit, + .init = init, + .uninit = uninit, + .query_formats = query_formats, + .activate = activate, + .priv_size = sizeof(IdentityContext), + .priv_class = &msad_class, + .inputs = identity_inputs, + .outputs = identity_outputs, + .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL | AVFILTER_FLAG_SLICE_THREADS, +}; + +#endif /* CONFIG_MSAD_FILTER */