From patchwork Fri Oct 8 09:57:35 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: 31000 Delivered-To: ffmpegpatchwork2@gmail.com Received: by 2002:a05:6602:2084:0:0:0:0 with SMTP id a4csp696746ioa; Fri, 8 Oct 2021 02:57:48 -0700 (PDT) X-Google-Smtp-Source: ABdhPJyysQPLdv5QogP9ouYYD4rDbZ6RY+QRf6dWphwiHGF0UUVCMKvbdtvoLfJQWNwAVGMxM9+q X-Received: by 2002:a05:6402:274b:: with SMTP id z11mr13933485edd.151.1633687067887; Fri, 08 Oct 2021 02:57:47 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1633687067; cv=none; d=google.com; s=arc-20160816; b=pZpjjiDg8k1vLgfjuUzT2VUw+XL3w+fYZp5mziC2AtaOsN46H+8rCOtk44MzTVXMGK DbkxsNmvS23OXJ84WexYLfZ+yzcc347vTix0NCo2kqjljNZVoan2O9HVKA/BTS4u+1eX EYZDyMC8pNyOnxm1P9XZSNo7DLh2Irw3p1/PXNJMN2uiMgYNBKzOf/IaGBeZ9IPCkMJs Qaatdk90N+hacxEwpsWkfyFeb1+107Pt1/DYNSDJqWaIWesbMK1Z1FoOPIEi9KNYNc0u 0gfTdKQX82jvd+3pj9eQPe8GjSFmLRHaOEDTz67mc2MFKN8naObG01w7YzoAGHJIe5gi x/lw== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:content-transfer-encoding:reply-to:list-subscribe :list-help:list-post:list-archive:list-unsubscribe:list-id :precedence:subject:mime-version:message-id:date:to:from :dkim-signature:delivered-to; bh=NpsjwpehnwadMh5hgftWTPVzTawbmzuuNSgTQDNgiOg=; b=X5BQnCHWJhKfI7U9rUlmNE/VcLdYXfTGhGqVpMPcvsvfyb9UOV89egtNshCHuQaBxi sXlcaPGl/mvBet4Ji+F+WwtA/pIavxDx+PGjS05w2GLjVNZZLFWJwZmXLZdqBgQgzBtz +VknC0rB1li/zOq8boM7O7ExvfPJhSRWuNOFc6VjAR0sjwCm4kWkQHSEgBld82/dK926 FQ9E0Mjh1N/V9CjCfTA60gcjyFODzUF5LiVxex6O5ppAz7tHSEVGg1LF52G+GD/8P7bd cW6J6fDglk9dR4MyGYePEghwzI6Ffpf3B4y0oS9aIyAm4D/JYwBOaWoqujSq6knxwULV T0hA== ARC-Authentication-Results: i=1; mx.google.com; dkim=neutral (body hash did not verify) header.i=@gmail.com header.s=20210112 header.b=fENvRlAt; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=QUARANTINE dis=NONE) header.from=gmail.com Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org. [79.124.17.100]) by mx.google.com with ESMTP id k23si2935924edj.602.2021.10.08.02.57.46; Fri, 08 Oct 2021 02:57:47 -0700 (PDT) Received-SPF: pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) client-ip=79.124.17.100; Authentication-Results: mx.google.com; dkim=neutral (body hash did not verify) header.i=@gmail.com header.s=20210112 header.b=fENvRlAt; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=QUARANTINE dis=NONE) header.from=gmail.com Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id C3EA968988E; Fri, 8 Oct 2021 12:57:42 +0300 (EEST) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from mail-wr1-f43.google.com (mail-wr1-f43.google.com [209.85.221.43]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id 3C642680CAB for ; Fri, 8 Oct 2021 12:57:36 +0300 (EEST) Received: by mail-wr1-f43.google.com with SMTP id k7so27861199wrd.13 for ; Fri, 08 Oct 2021 02:57:36 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112; h=from:to:subject:date:message-id:mime-version :content-transfer-encoding; bh=mP+N30LjY3Y8w4IQXIeTQO75lL7Lf4LsUn2avUPpl2o=; b=fENvRlAtA6xNtCtRcNydznonZhFypMQ9ju1StlWyJa6yMQ/wY7RNG0BzTir2Fz69K0 R5pEO2Pi8FlIsQKVICULzFTfV7Cl9CB4TbifxscxsXfZhC65pmiyYzb6T9NTYS6VE9xF ThwzS6dgreDvIdZ7g5I56MjfK7h9of/HrUEEhqe54ehtqM/pcHyL8f7s8BTJRsER2DOI HboNaCXmY88roKB62cU1xQdowoF8OLBCTGR47221i8P+DMfYBLcR1HSC0zliPeAhFbW2 tCpxpueau9CKZEIcI61xI+9HwMQcXBhF71oj6Bmp151gh6B+TiuWb1DJM6AJ/dZ3epsx GWBA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:from:to:subject:date:message-id:mime-version :content-transfer-encoding; bh=mP+N30LjY3Y8w4IQXIeTQO75lL7Lf4LsUn2avUPpl2o=; b=jwBR+sJCY568qwsWvvv3qSKg0MWk3/JrI9Y4zldNOvGJx5fKwu4T2UHr+5RyQurz01 yX6MBzQOaygj6jEYPmbODAj/e54XQl1eBZlEJ5SVO3czjCcqkxpqmqzeYAThsMupspxO zG9RPfWe+rkQaIX/Fflf/AhvYLYJ2qwzs9zF5Oq+QP1UOfHEKXENV20kGnrh6AmCpIVk Jr9TO6jSb1GNvBwJlg4fIkjrAEVFcy3+GJbWzFCbJcztrxxTOieJNqd61e0KWtWksmsn eu2ixb5nvZ6Q5WF8MUy/uKUxBQe8pHddCaHMCUv/6vNZz0OIuItEscMmfU0Dmx8Os/k4 SzLQ== X-Gm-Message-State: AOAM532CnYs6tdhpeE+CRnYnDIBMJIvy332d9nRba0BU4cCaOIkRIl7N RnZzMImHxcpcrqoFuyXOIMyq8boEFjU= X-Received: by 2002:adf:f584:: with SMTP id f4mr2838409wro.60.1633687055398; Fri, 08 Oct 2021 02:57:35 -0700 (PDT) Received: from localhost.localdomain ([95.168.118.138]) by smtp.gmail.com with ESMTPSA id q10sm2003272wmq.12.2021.10.08.02.57.33 for (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Fri, 08 Oct 2021 02:57:34 -0700 (PDT) From: Paul B Mahol To: ffmpeg-devel@ffmpeg.org Date: Fri, 8 Oct 2021 11:57:35 +0200 Message-Id: <20211008095736.211700-1-onemda@gmail.com> X-Mailer: git-send-email 2.33.0 MIME-Version: 1.0 Subject: [FFmpeg-devel] [PATCH 1/2] avfilter: add limitdiff video filter X-BeenThere: ffmpeg-devel@ffmpeg.org X-Mailman-Version: 2.1.29 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 Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" X-TUID: STt9VxYuaESA Signed-off-by: Paul B Mahol --- doc/filters.texi | 31 ++++ libavfilter/Makefile | 1 + libavfilter/allfilters.c | 1 + libavfilter/vf_limitdiff.c | 370 +++++++++++++++++++++++++++++++++++++ 4 files changed, 403 insertions(+) create mode 100644 libavfilter/vf_limitdiff.c diff --git a/doc/filters.texi b/doc/filters.texi index 633c64c0ea..2bd72b9303 100644 --- a/doc/filters.texi +++ b/doc/filters.texi @@ -14544,6 +14544,37 @@ ffmpeg -i main.mpg -i ref.mkv -lavfi "[0:v]settb=AVTB,setpts=PTS-STARTPTS[main]; @end example @end itemize +@section limitdiff +Apply limited difference filter using second and optionally third video stream. + +The filter accepts the following options: + +@table @option +@item threshold +Set the threshold to use when allowing certain differences between video streams. +Any absolute difference value lower or exact than this threshold will pick pixel components from +first video stream. + +@item elasticity +Set the elasticity of soft thresholding when processing video streams. +This value multiplied with first one sets second threshold. +Any absolute difference value greater or exact than second threshold will pick pixel components +from second video stream. For values between those two threshold +linear interpolation between first and second video stream will be used. + +@item reference +Enable the reference (third) video stream processing. By default is disabled. +If set, this video stream will be used for calculating absolute difference with first video +stream. + +@item planes +Specify which planes will be processed. Defaults to all available. +@end table + +@subsection Commands + +This filter supports the all above options as @ref{commands} except option @samp{reference}. + @section limiter Limits the pixel components values to the specified range [min, max]. diff --git a/libavfilter/Makefile b/libavfilter/Makefile index aec7da3f26..d22fcb574c 100644 --- a/libavfilter/Makefile +++ b/libavfilter/Makefile @@ -324,6 +324,7 @@ OBJS-$(CONFIG_LATENCY_FILTER) += f_latency.o OBJS-$(CONFIG_LENSCORRECTION_FILTER) += vf_lenscorrection.o OBJS-$(CONFIG_LENSFUN_FILTER) += vf_lensfun.o OBJS-$(CONFIG_LIBVMAF_FILTER) += vf_libvmaf.o framesync.o +OBJS-$(CONFIG_LIMITDIFF_FILTER) += vf_limitdiff.o framesync.o OBJS-$(CONFIG_LIMITER_FILTER) += vf_limiter.o OBJS-$(CONFIG_LOOP_FILTER) += f_loop.o OBJS-$(CONFIG_LUMAKEY_FILTER) += vf_lumakey.o diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c index 2da3ccaa59..95e34d97c6 100644 --- a/libavfilter/allfilters.c +++ b/libavfilter/allfilters.c @@ -309,6 +309,7 @@ extern const AVFilter ff_vf_latency; extern const AVFilter ff_vf_lenscorrection; extern const AVFilter ff_vf_lensfun; extern const AVFilter ff_vf_libvmaf; +extern const AVFilter ff_vf_limitdiff; extern const AVFilter ff_vf_limiter; extern const AVFilter ff_vf_loop; extern const AVFilter ff_vf_lumakey; diff --git a/libavfilter/vf_limitdiff.c b/libavfilter/vf_limitdiff.c new file mode 100644 index 0000000000..d2688c39f4 --- /dev/null +++ b/libavfilter/vf_limitdiff.c @@ -0,0 +1,370 @@ +/* + * Copyright (c) 2021 Paul B Mahol + * + * 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/imgutils.h" +#include "libavutil/pixdesc.h" +#include "libavutil/opt.h" +#include "avfilter.h" +#include "formats.h" +#include "internal.h" +#include "video.h" +#include "framesync.h" + +typedef struct ThreadData { + AVFrame *filtered, *source, *reference, *dst; +} ThreadData; + +typedef struct LimitDiffContext { + const AVClass *class; + + float threshold; + float elasticity; + int reference; + int planes; + + int thr1, thr2; + + int linesize[4]; + int planewidth[4], planeheight[4]; + int nb_planes; + int depth; + FFFrameSync fs; + + void (*limitdiff)(const uint8_t *filtered, uint8_t *dst, + const uint8_t *source, const uint8_t *reference, + int thr1, int thr2, int w, int depth); +} LimitDiffContext; + +#define OFFSET(x) offsetof(LimitDiffContext, x) +#define TFLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_RUNTIME_PARAM +#define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM + +static const AVOption limitdiff_options[] = { + { "threshold", "set the threshold", OFFSET(threshold), AV_OPT_TYPE_FLOAT, {.dbl=1/255.f}, 0, 1, TFLAGS }, + { "elasticity", "set the elasticity", OFFSET(elasticity), AV_OPT_TYPE_FLOAT, {.dbl=2.f}, 0, 10, TFLAGS }, + { "reference", "enable reference stream", OFFSET(reference), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS }, + { "planes", "set the planes to filter", OFFSET(planes), AV_OPT_TYPE_INT, {.i64=0xF}, 0, 0xF, TFLAGS }, + { NULL } +}; + +static const enum AVPixelFormat pix_fmts[] = { + AV_PIX_FMT_YUVA444P, AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV440P, + AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUVJ440P, + AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUV420P, + AV_PIX_FMT_YUVJ422P, AV_PIX_FMT_YUVJ420P, + AV_PIX_FMT_YUVJ411P, AV_PIX_FMT_YUV411P, AV_PIX_FMT_YUV410P, + AV_PIX_FMT_YUV420P9, AV_PIX_FMT_YUV422P9, AV_PIX_FMT_YUV444P9, + AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV444P10, + AV_PIX_FMT_YUV420P12, AV_PIX_FMT_YUV422P12, AV_PIX_FMT_YUV444P12, AV_PIX_FMT_YUV440P12, + AV_PIX_FMT_YUV420P14, AV_PIX_FMT_YUV422P14, AV_PIX_FMT_YUV444P14, + AV_PIX_FMT_YUV420P16, AV_PIX_FMT_YUV422P16, AV_PIX_FMT_YUV444P16, + AV_PIX_FMT_YUVA420P9, AV_PIX_FMT_YUVA422P9, AV_PIX_FMT_YUVA444P9, + AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_YUVA444P10, + AV_PIX_FMT_YUVA422P12, AV_PIX_FMT_YUVA444P12, + AV_PIX_FMT_YUVA420P16, AV_PIX_FMT_YUVA422P16, AV_PIX_FMT_YUVA444P16, + AV_PIX_FMT_GBRP, AV_PIX_FMT_GBRP9, AV_PIX_FMT_GBRP10, + AV_PIX_FMT_GBRP12, AV_PIX_FMT_GBRP14, AV_PIX_FMT_GBRP16, + AV_PIX_FMT_GBRAP, AV_PIX_FMT_GBRAP10, AV_PIX_FMT_GBRAP12, AV_PIX_FMT_GBRAP16, + AV_PIX_FMT_GRAY8, AV_PIX_FMT_GRAY9, AV_PIX_FMT_GRAY10, AV_PIX_FMT_GRAY12, AV_PIX_FMT_GRAY14, AV_PIX_FMT_GRAY16, + AV_PIX_FMT_NONE +}; + +static void limitdiff8(const uint8_t *filtered, uint8_t *dst, + const uint8_t *source, const uint8_t *reference, + int thr1, int thr2, int w, int depth) +{ + for (int x = 0; x < w; x++) { + const int diff = filtered[x] - source[x]; + const int diff_ref = FFABS(filtered[x] - reference[x]); + + if (diff_ref <= thr1) + dst[x] = filtered[x]; + else if (diff_ref >= thr2) + dst[x] = source[x]; + else + dst[x] = av_clip_uint8(source[x] + diff * (thr2 - diff_ref) / (thr2 - thr1)); + } +} + +static void limitdiff16(const uint8_t *ffiltered, uint8_t *ddst, + const uint8_t *ssource, const uint8_t *rreference, + int thr1, int thr2, int w, int depth) +{ + const uint16_t *source = (const uint16_t *)ssource; + const uint16_t *filtered = (const uint16_t *)ffiltered; + const uint16_t *reference = (const uint16_t *)rreference; + uint16_t *dst = (uint16_t *)ddst; + + for (int x = 0; x < w; x++) { + const int diff = filtered[x] - source[x]; + const int diff_ref = FFABS(filtered[x] - reference[x]); + + if (diff_ref <= thr1) + dst[x] = filtered[x]; + else if (diff_ref >= thr2) + dst[x] = source[x]; + else + dst[x] = av_clip_uintp2_c(source[x] + diff * (thr2 - diff_ref) / (thr2 - thr1), depth); + } +} + +static int config_input(AVFilterLink *inlink) +{ + AVFilterContext *ctx = inlink->dst; + LimitDiffContext *s = ctx->priv; + const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format); + int vsub, hsub, ret; + + s->nb_planes = av_pix_fmt_count_planes(inlink->format); + + if ((ret = av_image_fill_linesizes(s->linesize, inlink->format, inlink->w)) < 0) + return ret; + + hsub = desc->log2_chroma_w; + vsub = desc->log2_chroma_h; + s->planeheight[1] = s->planeheight[2] = AV_CEIL_RSHIFT(inlink->h, vsub); + s->planeheight[0] = s->planeheight[3] = inlink->h; + s->planewidth[1] = s->planewidth[2] = AV_CEIL_RSHIFT(inlink->w, hsub); + s->planewidth[0] = s->planewidth[3] = inlink->w; + + s->depth = desc->comp[0].depth; + s->thr1 = s->threshold * ((1 << s->depth) - 1); + s->thr2 = s->thr1 * s->elasticity; + + if (desc->comp[0].depth == 8) + s->limitdiff = limitdiff8; + else + s->limitdiff = limitdiff16; + + return 0; +} + +static int limitdiff_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) +{ + LimitDiffContext *s = ctx->priv; + const int depth = s->depth; + ThreadData *td = arg; + + for (int p = 0; p < s->nb_planes; p++) { + const ptrdiff_t filtered_linesize = td->filtered->linesize[p]; + const ptrdiff_t source_linesize = td->source->linesize[p]; + const ptrdiff_t reference_linesize = td->reference->linesize[p]; + const ptrdiff_t dst_linesize = td->dst->linesize[p]; + const int thr1 = s->thr1; + const int thr2 = s->thr2; + const int w = s->planewidth[p]; + const int h = s->planeheight[p]; + const int slice_start = (h * jobnr) / nb_jobs; + const int slice_end = (h * (jobnr+1)) / nb_jobs; + const uint8_t *filtered = td->filtered->data[p] + slice_start * filtered_linesize; + const uint8_t *source = td->source->data[p] + slice_start * source_linesize; + const uint8_t *reference = td->reference->data[p] + slice_start * reference_linesize; + uint8_t *dst = td->dst->data[p] + slice_start * dst_linesize; + + if (!((1 << p) & s->planes)) { + av_image_copy_plane(dst, dst_linesize, filtered, filtered_linesize, + s->linesize[p], slice_end - slice_start); + continue; + } + + for (int y = slice_start; y < slice_end; y++) { + s->limitdiff(filtered, dst, source, reference, thr1, thr2, w, depth); + + dst += dst_linesize; + filtered += filtered_linesize; + source += source_linesize; + reference += reference_linesize; + } + } + + return 0; +} + +static int process_frame(FFFrameSync *fs) +{ + AVFilterContext *ctx = fs->parent; + LimitDiffContext *s = fs->opaque; + AVFilterLink *outlink = ctx->outputs[0]; + AVFrame *out, *filtered, *source, *reference = NULL; + int ret; + + if ((ret = ff_framesync_get_frame(&s->fs, 0, &filtered, 0)) < 0 || + (ret = ff_framesync_get_frame(&s->fs, 1, &source, 0)) < 0) + return ret; + if (s->reference) { + if ((ret = ff_framesync_get_frame(&s->fs, 2, &reference, 0)) < 0) + return ret; + } + + if (ctx->is_disabled) { + out = av_frame_clone(filtered); + if (!out) + return AVERROR(ENOMEM); + } else { + ThreadData td; + + out = ff_get_video_buffer(outlink, outlink->w, outlink->h); + if (!out) + return AVERROR(ENOMEM); + av_frame_copy_props(out, filtered); + + td.filtered = filtered; + td.source = source; + td.reference = reference ? reference : source; + td.dst = out; + + ff_filter_execute(ctx, limitdiff_slice, &td, NULL, + FFMIN(s->planeheight[0], ff_filter_get_nb_threads(ctx))); + } + out->pts = av_rescale_q(s->fs.pts, s->fs.time_base, outlink->time_base); + + return ff_filter_frame(outlink, out); +} + +static int config_output(AVFilterLink *outlink) +{ + AVFilterContext *ctx = outlink->src; + LimitDiffContext *s = ctx->priv; + AVFilterLink *filtered = ctx->inputs[0]; + AVFilterLink *source = ctx->inputs[1]; + FFFrameSyncIn *in; + int ret; + + if (filtered->w != source->w || filtered->h != source->h) { + av_log(ctx, AV_LOG_ERROR, "First input link %s parameters " + "(size %dx%d) do not match the corresponding " + "second input link %s parameters (%dx%d)\n", + ctx->input_pads[0].name, filtered->w, filtered->h, + ctx->input_pads[1].name, source->w, source->h); + return AVERROR(EINVAL); + } + + if (s->reference) { + AVFilterLink *reference = ctx->inputs[2]; + + if (filtered->w != reference->w || filtered->h != reference->h) { + av_log(ctx, AV_LOG_ERROR, "First input link %s parameters " + "(size %dx%d) do not match the corresponding " + "third input link %s parameters (%dx%d)\n", + ctx->input_pads[0].name, filtered->w, filtered->h, + ctx->input_pads[1].name, reference->w, reference->h); + return AVERROR(EINVAL); + } + } + + outlink->w = filtered->w; + outlink->h = filtered->h; + outlink->sample_aspect_ratio = filtered->sample_aspect_ratio; + outlink->frame_rate = filtered->frame_rate; + + if ((ret = ff_framesync_init(&s->fs, ctx, 2 + !!s->reference)) < 0) + return ret; + + in = s->fs.in; + in[0].time_base = filtered->time_base; + in[1].time_base = source->time_base; + if (s->reference) + in[2].time_base = ctx->inputs[2]->time_base; + in[0].sync = 1; + in[0].before = EXT_STOP; + in[0].after = EXT_INFINITY; + in[1].sync = 1; + in[1].before = EXT_STOP; + in[1].after = EXT_INFINITY; + if (s->reference) { + in[2].sync = 1; + in[2].before = EXT_STOP; + in[2].after = EXT_INFINITY; + } + s->fs.opaque = s; + s->fs.on_event = process_frame; + + ret = ff_framesync_configure(&s->fs); + outlink->time_base = s->fs.time_base; + + return ret; +} + +static int activate(AVFilterContext *ctx) +{ + LimitDiffContext *s = ctx->priv; + return ff_framesync_activate(&s->fs); +} + +static av_cold int init(AVFilterContext *ctx) +{ + const LimitDiffContext *s = ctx->priv; + AVFilterPad pad = { + .name = "filtered", + .type = AVMEDIA_TYPE_VIDEO, + .config_props = config_input, + }; + int ret; + + if ((ret = ff_append_inpad(ctx, &pad)) < 0) + return ret; + + pad.name = "source"; + pad.config_props = NULL; + if ((ret = ff_append_inpad(ctx, &pad)) < 0) + return ret; + + if (s->reference) { + pad.name = "reference"; + pad.config_props = NULL; + if ((ret = ff_append_inpad(ctx, &pad)) < 0) + return ret; + } + + return 0; +} + +static av_cold void uninit(AVFilterContext *ctx) +{ + LimitDiffContext *s = ctx->priv; + + ff_framesync_uninit(&s->fs); +} + +static const AVFilterPad limitdiff_outputs[] = { + { + .name = "default", + .type = AVMEDIA_TYPE_VIDEO, + .config_props = config_output, + }, +}; + +AVFILTER_DEFINE_CLASS(limitdiff); + +const AVFilter ff_vf_limitdiff = { + .name = "limitdiff", + .description = NULL_IF_CONFIG_SMALL("Apply filtering with limiting difference."), + .priv_class = &limitdiff_class, + .priv_size = sizeof(LimitDiffContext), + .init = init, + .uninit = uninit, + .activate = activate, + FILTER_OUTPUTS(limitdiff_outputs), + FILTER_PIXFMTS_ARRAY(pix_fmts), + .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL | + AVFILTER_FLAG_SLICE_THREADS | + AVFILTER_FLAG_DYNAMIC_INPUTS, + .process_command = ff_filter_process_command, +}; From patchwork Fri Oct 8 09:57:36 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: 30999 Delivered-To: ffmpegpatchwork2@gmail.com Received: by 2002:a05:6602:2084:0:0:0:0 with SMTP id a4csp696888ioa; Fri, 8 Oct 2021 02:57:59 -0700 (PDT) X-Google-Smtp-Source: ABdhPJyju8kcYH+RjMVNt6aQ5ZFAm+7AzcLUf9LjHlXs5zYVRL3esY2jeNebjKPePoAn3ZlKS6Pg X-Received: by 2002:a05:6402:2052:: with SMTP id bc18mr13897169edb.190.1633687078889; Fri, 08 Oct 2021 02:57:58 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1633687078; cv=none; d=google.com; s=arc-20160816; b=DjfBZRIAF3+pSTu3JEKYR71tfNrNpvJnk4akGRDqNljid4nEwVNSA7xMNx8LsK6dMy 66iqPgVwU8zHb1mnaw5WMjKOPJRbP9bk+eclRpy6d/h6JjjMR88MIPY9CiIe7JCaGkRA fiYXo6ck7YUTAtD9nvqVCbC3H94ammbEdRO8IoaqpKt2KEdPwAy7LWryX2Gg/Z1P0rZT pZhwECPa74hESfUPoAscoAw3Muc9p/5s6suoPyZdxXEFKVnMf+mjlme0uMao1Vk6iMry c/PEhPzn8Dir0LI8eTtHjV0PcT1gSl6UvpVabcXNI1gOsZbQaScEZg2wUltdpqlbrq/Q OHbQ== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:content-transfer-encoding:reply-to:list-subscribe :list-help:list-post:list-archive:list-unsubscribe:list-id :precedence:subject:mime-version:references:in-reply-to:message-id :date:to:from:dkim-signature:delivered-to; bh=dJFaYLT8JmJPxF40O/U8/en8zSy9ArkrDW4jLKqDdMo=; b=QVPSIIIsD6rW14ykDIuJDZIQx00UhGzOc5OjzKArONh3qx9oQlIoqWILe8mz4v6BDm 0nIelc8mSGV/G83AEZ3+wxhoLJ/yTrkwKh0UK8jmFB/abgyEfXKaEqTuspjlMD4KlyP2 SiE2oPjqFc/58rvL7yLU5tIA9ptIeX4JqQicLiTPIV7Fr4WUV1HNtWk+dasqr7+Z68lJ pbTEeE2c2Ed3sm2p9wNUuRsEAptG1RAvPDdAz2U1fIjyfdXQdMcpU6iPUD6GrJJSpbhh cfM/NjKgDakZT0k44wwebGcGD3GeynMD8jWaLYF7x3oqHHR6ePzKQANpqpY1xg2smN/3 zz3w== ARC-Authentication-Results: i=1; mx.google.com; dkim=neutral (body hash did not verify) header.i=@gmail.com header.s=20210112 header.b=GzL6O+ig; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=QUARANTINE dis=NONE) header.from=gmail.com Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org. [79.124.17.100]) by mx.google.com with ESMTP id hp5si4846852ejc.447.2021.10.08.02.57.58; Fri, 08 Oct 2021 02:57:58 -0700 (PDT) Received-SPF: pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) client-ip=79.124.17.100; Authentication-Results: mx.google.com; dkim=neutral (body hash did not verify) header.i=@gmail.com header.s=20210112 header.b=GzL6O+ig; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=QUARANTINE dis=NONE) header.from=gmail.com Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id D876B689E56; Fri, 8 Oct 2021 12:57:43 +0300 (EEST) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from mail-wr1-f52.google.com (mail-wr1-f52.google.com [209.85.221.52]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id 092306881B1 for ; Fri, 8 Oct 2021 12:57:37 +0300 (EEST) Received: by mail-wr1-f52.google.com with SMTP id k7so27861294wrd.13 for ; Fri, 08 Oct 2021 02:57:36 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112; h=from:to:subject:date:message-id:in-reply-to:references:mime-version :content-transfer-encoding; bh=tEIeu4dEA77lsfDMyY/Y2XGd6xwIDjuwonYejcg1kPA=; b=GzL6O+igqVr43odNLKhKBmBtKtzzU0r/ynJ+SqrNEV988KRwaPip+T0TJ2OwZZ91S9 83OqTcS72vYEWfz0/ZE86oBiayglrUj678k/x90Woz/jlfoKbKXAYl+5OsmbD4qOSvcU jASLHKgJyNcfxw4jJJn/D7bVkSdH4iXbaLsHPpeI2/ezRZiGSaP6Z589UcD6K4B7jLut dBnM91qMtbRuPS7PC6HdPKXb6joZ2oJtawXCHu/VbvkETGsPow/PDW6LhxzPcu9okA97 RiDk2feZnUchKSuezUJMMtRYEVUJp3qudgtZUGRYVFWgfWFyYe+h+CgvzubekE+/loXt d+vA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:from:to:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=tEIeu4dEA77lsfDMyY/Y2XGd6xwIDjuwonYejcg1kPA=; b=i+0Z3nAg9+YvCLkuYLWt9VwuGqm/YCWJsp2m7ZHLj6a1VX/Xlgq4TdPmnBzw/SokLE 4BbS0olqIfm5jQ4uS6TrSDbCSEy2jawHgLcaMyhjHDTk1/WzYeQOt1+fTQ7kb/qIPpH1 sfgwr9WUHJNLsAO0ClxY4qKv6uw37B71H4BMKH0Qgc3TNBQtB+TltivVkwVULySWyOWz unEAaDYb4i/M9hCW7uIgvmUCNca5pjI5mYpamQ4lZk40S2Lu/5Ziy/s4V3fMbSU6Avq4 MSQT74AgWhHT1qdwDpMJWma+6j9+hwp3lHXVEAskzAQG2HqYq30w8CT6gSpsRhzHX0LA zDXA== X-Gm-Message-State: AOAM5319AX0/EsXCX51qDMbhEGI6QgpB0AmCiCy6ns/OVNNOZ3IUg3bs +IYPsZ5VGkNjCI8IMhwIDngGTrq//+4= X-Received: by 2002:a1c:2209:: with SMTP id i9mr2494139wmi.20.1633687056359; Fri, 08 Oct 2021 02:57:36 -0700 (PDT) Received: from localhost.localdomain ([95.168.118.138]) by smtp.gmail.com with ESMTPSA id q10sm2003272wmq.12.2021.10.08.02.57.35 for (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Fri, 08 Oct 2021 02:57:36 -0700 (PDT) From: Paul B Mahol To: ffmpeg-devel@ffmpeg.org Date: Fri, 8 Oct 2021 11:57:36 +0200 Message-Id: <20211008095736.211700-2-onemda@gmail.com> X-Mailer: git-send-email 2.33.0 In-Reply-To: <20211008095736.211700-1-onemda@gmail.com> References: <20211008095736.211700-1-onemda@gmail.com> MIME-Version: 1.0 Subject: [FFmpeg-devel] [PATCH 2/2] avfilter: add xcorrelate video filter X-BeenThere: ffmpeg-devel@ffmpeg.org X-Mailman-Version: 2.1.29 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 Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" X-TUID: 3j5ZCoHWEzKO Signed-off-by: Paul B Mahol --- doc/filters.texi | 18 ++ libavfilter/Makefile | 1 + libavfilter/allfilters.c | 1 + libavfilter/vf_convolve.c | 370 +++++++++++++++++++++++++++++++++----- 4 files changed, 346 insertions(+), 44 deletions(-) diff --git a/doc/filters.texi b/doc/filters.texi index 2bd72b9303..7266bc0ddb 100644 --- a/doc/filters.texi +++ b/doc/filters.texi @@ -22611,6 +22611,24 @@ Set the scaling dimension: @code{2} for @code{2xBR}, @code{3} for Default is @code{3}. @end table +@section xcorrelate +Apply cross-correlation between first and second input video stream. + +Second input video stream dimensions must be lower than first input video stream. + +The filter accepts the following options: + +@table @option +@item planes +Set which planes to process. + +@item secondary +Set which secondary video frames will be processed from second input video stream, +can be @var{first} or @var{all}. Default is @var{all}. +@end table + +The @code{xcorrelate} filter also supports the @ref{framesync} options. + @section xfade Apply cross fade from one input video stream to another input video stream. diff --git a/libavfilter/Makefile b/libavfilter/Makefile index d22fcb574c..da2c8b7a5e 100644 --- a/libavfilter/Makefile +++ b/libavfilter/Makefile @@ -501,6 +501,7 @@ OBJS-$(CONFIG_W3FDIF_FILTER) += vf_w3fdif.o OBJS-$(CONFIG_WAVEFORM_FILTER) += vf_waveform.o OBJS-$(CONFIG_WEAVE_FILTER) += vf_weave.o OBJS-$(CONFIG_XBR_FILTER) += vf_xbr.o +OBJS-$(CONFIG_XCORRELATE_FILTER) += vf_convolve.o framesync.o OBJS-$(CONFIG_XFADE_FILTER) += vf_xfade.o OBJS-$(CONFIG_XFADE_OPENCL_FILTER) += vf_xfade_opencl.o opencl.o opencl/xfade.o OBJS-$(CONFIG_XMEDIAN_FILTER) += vf_xmedian.o framesync.o diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c index 95e34d97c6..81e1b0965b 100644 --- a/libavfilter/allfilters.c +++ b/libavfilter/allfilters.c @@ -478,6 +478,7 @@ extern const AVFilter ff_vf_w3fdif; extern const AVFilter ff_vf_waveform; extern const AVFilter ff_vf_weave; extern const AVFilter ff_vf_xbr; +extern const AVFilter ff_vf_xcorrelate; extern const AVFilter ff_vf_xfade; extern const AVFilter ff_vf_xfade_opencl; extern const AVFilter ff_vf_xmedian; diff --git a/libavfilter/vf_convolve.c b/libavfilter/vf_convolve.c index 9d506d49dd..55afb582b4 100644 --- a/libavfilter/vf_convolve.c +++ b/libavfilter/vf_convolve.c @@ -47,6 +47,12 @@ typedef struct ConvolveContext { int planewidth[4]; int planeheight[4]; + int primarywidth[4]; + int primaryheight[4]; + + int secondarywidth[4]; + int secondaryheight[4]; + AVComplexFloat *fft_hdata_in[4]; AVComplexFloat *fft_vdata_in[4]; AVComplexFloat *fft_hdata_out[4]; @@ -63,6 +69,13 @@ typedef struct ConvolveContext { int nb_planes; int got_impulse[4]; + void (*get_input)(struct ConvolveContext *s, AVComplexFloat *fft_hdata, + AVFrame *in, int w, int h, int n, int plane, float scale); + + void (*get_output)(struct ConvolveContext *s, AVComplexFloat *input, AVFrame *out, + int w, int h, int n, int plane, float scale); + void (*prepare_impulse)(AVFilterContext *ctx, AVFrame *impulsepic, int plane); + int (*filter)(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs); } ConvolveContext; @@ -99,21 +112,22 @@ static const enum AVPixelFormat pixel_fmts_fftfilt[] = { AV_PIX_FMT_NONE }; -static int config_input_main(AVFilterLink *inlink) +static int config_input(AVFilterLink *inlink) { ConvolveContext *s = inlink->dst->priv; const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format); - int i; + const int w = inlink->w; + const int h = inlink->h; - 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->planewidth[1] = s->planewidth[2] = AV_CEIL_RSHIFT(w, desc->log2_chroma_w); + s->planewidth[0] = s->planewidth[3] = w; + s->planeheight[1] = s->planeheight[2] = AV_CEIL_RSHIFT(h, desc->log2_chroma_h); + s->planeheight[0] = s->planeheight[3] = h; s->nb_planes = desc->nb_components; s->depth = desc->comp[0].depth; - for (i = 0; i < s->nb_planes; i++) { + for (int i = 0; i < s->nb_planes; i++) { int w = s->planewidth[i]; int h = s->planeheight[i]; int n = FFMAX(w, h); @@ -186,6 +200,98 @@ static int fft_horizontal(AVFilterContext *ctx, void *arg, int jobnr, int nb_job return 0; } +#define SQR(x) ((x) * (x)) + +static void get_zeropadded_input(ConvolveContext *s, + AVComplexFloat *fft_hdata, + AVFrame *in, int w, int h, + int n, int plane, float scale) +{ + float sum = 0.f; + float mean, dev; + int y, x; + + if (s->depth == 8) { + for (y = 0; y < h; y++) { + const uint8_t *src = in->data[plane] + in->linesize[plane] * y; + + for (x = 0; x < w; x++) + sum += src[x]; + } + + mean = sum / (w * h); + sum = 0.f; + for (y = 0; y < h; y++) { + const uint8_t *src = in->data[plane] + in->linesize[plane] * y; + + for (x = 0; x < w; x++) + sum += SQR(src[x] - mean); + } + + dev = sqrtf(sum / (w * h)); + scale /= dev; + for (y = 0; y < h; y++) { + const uint8_t *src = in->data[plane] + in->linesize[plane] * y; + + for (x = 0; x < w; x++) { + fft_hdata[y * n + x].re = (src[x] - mean) * scale; + fft_hdata[y * n + x].im = 0; + } + + for (x = w; x < n; x++) { + fft_hdata[y * n + x].re = 0; + fft_hdata[y * n + x].im = 0; + } + } + + for (y = h; y < n; y++) { + for (x = 0; x < n; x++) { + fft_hdata[y * n + x].re = 0; + fft_hdata[y * n + x].im = 0; + } + } + } else { + for (y = 0; y < h; y++) { + const uint16_t *src = (const uint16_t *)(in->data[plane] + in->linesize[plane] * y); + + for (x = 0; x < w; x++) + sum += src[x]; + } + + mean = sum / (w * h); + sum = 0.f; + for (y = 0; y < h; y++) { + const uint16_t *src = (const uint16_t *)(in->data[plane] + in->linesize[plane] * y); + + for (x = 0; x < w; x++) + sum += SQR(src[x] - mean); + } + + dev = sqrtf(sum / (w * h)); + scale /= dev; + for (y = 0; y < h; y++) { + const uint16_t *src = (const uint16_t *)(in->data[plane] + in->linesize[plane] * y); + + for (x = 0; x < w; x++) { + fft_hdata[y * n + x].re = (src[x] - mean) * scale; + fft_hdata[y * n + x].im = 0; + } + + for (x = w; x < n; x++) { + fft_hdata[y * n + x].re = 0; + fft_hdata[y * n + x].im = 0; + } + } + + for (y = h; y < n; y++) { + for (x = 0; x < n; x++) { + fft_hdata[y * n + x].re = 0; + fft_hdata[y * n + x].im = 0; + } + } + } +} + static void get_input(ConvolveContext *s, AVComplexFloat *fft_hdata, AVFrame *in, int w, int h, int n, int plane, float scale) { @@ -330,6 +436,27 @@ static int ifft_horizontal(AVFilterContext *ctx, void *arg, int jobnr, int nb_jo return 0; } +static void get_xoutput(ConvolveContext *s, AVComplexFloat *input, AVFrame *out, + int w, int h, int n, int plane, float scale) +{ + const int imax = (1 << s->depth) - 1; + + scale *= imax * 16; + if (s->depth == 8) { + for (int y = 0; y < h; y++) { + uint8_t *dst = out->data[plane] + y * out->linesize[plane]; + for (int x = 0; x < w; x++) + dst[x] = av_clip_uint8(input[y * n + x].re * scale); + } + } else { + for (int y = 0; y < h; y++) { + uint16_t *dst = (uint16_t *)(out->data[plane] + y * out->linesize[plane]); + for (int x = 0; x < w; x++) + dst[x] = av_clip(input[y * n + x].re * scale, 0, imax); + } + } +} + static void get_output(ConvolveContext *s, AVComplexFloat *input, AVFrame *out, int w, int h, int n, int plane, float scale) { @@ -414,6 +541,35 @@ static int complex_multiply(AVFilterContext *ctx, void *arg, int jobnr, int nb_j return 0; } +static int complex_xcorrelate(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) +{ + ThreadData *td = arg; + AVComplexFloat *input = td->hdata_in; + AVComplexFloat *filter = td->vdata_in; + const int n = td->n; + const float scale = 1.f / (n * n); + int start = (n * jobnr) / nb_jobs; + int end = (n * (jobnr+1)) / nb_jobs; + + for (int y = start; y < end; y++) { + int yn = y * n; + + for (int x = 0; x < n; x++) { + float re, im, ire, iim; + + re = input[yn + x].re; + im = input[yn + x].im; + ire = filter[yn + x].re * scale; + iim = -filter[yn + x].im * scale; + + input[yn + x].re = ire * re - iim * im; + input[yn + x].im = iim * re + ire * im; + } + } + + return 0; +} + static int complex_divide(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) { ConvolveContext *s = ctx->priv; @@ -446,13 +602,82 @@ static int complex_divide(AVFilterContext *ctx, void *arg, int jobnr, int nb_job return 0; } +static void prepare_impulse(AVFilterContext *ctx, AVFrame *impulsepic, int plane) +{ + ConvolveContext *s = ctx->priv; + const int n = s->fft_len[plane]; + const int w = s->secondarywidth[plane]; + const int h = s->secondaryheight[plane]; + ThreadData td; + float total = 0; + + if (s->depth == 8) { + for (int y = 0; y < h; y++) { + const uint8_t *src = (const uint8_t *)(impulsepic->data[plane] + y * impulsepic->linesize[plane]) ; + for (int x = 0; x < w; x++) { + total += src[x]; + } + } + } else { + for (int y = 0; y < h; y++) { + const uint16_t *src = (const uint16_t *)(impulsepic->data[plane] + y * impulsepic->linesize[plane]) ; + for (int x = 0; x < w; x++) { + total += src[x]; + } + } + } + total = FFMAX(1, total); + + s->get_input(s, s->fft_hdata_impulse_in[plane], impulsepic, w, h, n, plane, 1.f / total); + + td.n = n; + td.plane = plane; + td.hdata_in = s->fft_hdata_impulse_in[plane]; + td.vdata_in = s->fft_vdata_impulse_in[plane]; + td.hdata_out = s->fft_hdata_impulse_out[plane]; + td.vdata_out = s->fft_vdata_impulse_out[plane]; + + ff_filter_execute(ctx, fft_horizontal, &td, NULL, + FFMIN3(MAX_THREADS, n, ff_filter_get_nb_threads(ctx))); + ff_filter_execute(ctx, fft_vertical, &td, NULL, + FFMIN3(MAX_THREADS, n, ff_filter_get_nb_threads(ctx))); + + s->got_impulse[plane] = 1; +} + +static void prepare_secondary(AVFilterContext *ctx, AVFrame *secondary, int plane) +{ + ConvolveContext *s = ctx->priv; + const int n = s->fft_len[plane]; + ThreadData td; + + s->get_input(s, s->fft_hdata_impulse_in[plane], secondary, + s->secondarywidth[plane], + s->secondaryheight[plane], + n, plane, 1.f); + + td.n = n; + td.plane = plane; + td.hdata_in = s->fft_hdata_impulse_in[plane]; + td.vdata_in = s->fft_vdata_impulse_in[plane]; + td.hdata_out = s->fft_hdata_impulse_out[plane]; + td.vdata_out = s->fft_vdata_impulse_out[plane]; + + ff_filter_execute(ctx, fft_horizontal, &td, NULL, + FFMIN3(MAX_THREADS, n, ff_filter_get_nb_threads(ctx))); + ff_filter_execute(ctx, fft_vertical, &td, NULL, + FFMIN3(MAX_THREADS, n, ff_filter_get_nb_threads(ctx))); + + s->got_impulse[plane] = 1; +} + static int do_convolve(FFFrameSync *fs) { AVFilterContext *ctx = fs->parent; AVFilterLink *outlink = ctx->outputs[0]; ConvolveContext *s = ctx->priv; AVFrame *mainpic = NULL, *impulsepic = NULL; - int ret, y, x, plane; + int ret, plane; ret = ff_framesync_dualinput_get(fs, &mainpic, &impulsepic); if (ret < 0) @@ -464,9 +689,10 @@ static int do_convolve(FFFrameSync *fs) AVComplexFloat *filter = s->fft_vdata_impulse_out[plane]; AVComplexFloat *input = s->fft_vdata_out[plane]; const int n = s->fft_len[plane]; - const int w = s->planewidth[plane]; - const int h = s->planeheight[plane]; - float total = 0; + const int w = s->primarywidth[plane]; + const int h = s->primaryheight[plane]; + const int ow = s->planewidth[plane]; + const int oh = s->planeheight[plane]; ThreadData td; if (!(s->planes & (1 << plane))) { @@ -474,7 +700,7 @@ static int do_convolve(FFFrameSync *fs) } td.plane = plane, td.n = n; - get_input(s, s->fft_hdata_in[plane], mainpic, w, h, n, plane, 1.f); + s->get_input(s, s->fft_hdata_in[plane], mainpic, w, h, n, plane, 1.f); td.hdata_in = s->fft_hdata_in[plane]; td.vdata_in = s->fft_vdata_in[plane]; @@ -487,36 +713,7 @@ static int do_convolve(FFFrameSync *fs) FFMIN3(MAX_THREADS, n, ff_filter_get_nb_threads(ctx))); if ((!s->impulse && !s->got_impulse[plane]) || s->impulse) { - if (s->depth == 8) { - for (y = 0; y < h; y++) { - const uint8_t *src = (const uint8_t *)(impulsepic->data[plane] + y * impulsepic->linesize[plane]) ; - for (x = 0; x < w; x++) { - total += src[x]; - } - } - } else { - for (y = 0; y < h; y++) { - const uint16_t *src = (const uint16_t *)(impulsepic->data[plane] + y * impulsepic->linesize[plane]) ; - for (x = 0; x < w; x++) { - total += src[x]; - } - } - } - total = FFMAX(1, total); - - get_input(s, s->fft_hdata_impulse_in[plane], impulsepic, w, h, n, plane, 1.f / total); - - td.hdata_in = s->fft_hdata_impulse_in[plane]; - td.vdata_in = s->fft_vdata_impulse_in[plane]; - td.hdata_out = s->fft_hdata_impulse_out[plane]; - td.vdata_out = s->fft_vdata_impulse_out[plane]; - - ff_filter_execute(ctx, fft_horizontal, &td, NULL, - FFMIN3(MAX_THREADS, n, ff_filter_get_nb_threads(ctx))); - ff_filter_execute(ctx, fft_vertical, &td, NULL, - FFMIN3(MAX_THREADS, n, ff_filter_get_nb_threads(ctx))); - - s->got_impulse[plane] = 1; + s->prepare_impulse(ctx, impulsepic, plane); } td.hdata_in = input; @@ -539,7 +736,7 @@ static int do_convolve(FFFrameSync *fs) ff_filter_execute(ctx, ifft_horizontal, &td, NULL, FFMIN3(MAX_THREADS, n, ff_filter_get_nb_threads(ctx))); - get_output(s, s->fft_hdata_out[plane], mainpic, w, h, n, plane, 1.f / (n * n)); + s->get_output(s, s->fft_hdata_out[plane], mainpic, ow, oh, n, plane, 1.f / (n * n)); } return ff_filter_frame(outlink, mainpic); @@ -547,11 +744,23 @@ static int do_convolve(FFFrameSync *fs) static int config_output(AVFilterLink *outlink) { + const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(outlink->format); AVFilterContext *ctx = outlink->src; ConvolveContext *s = ctx->priv; AVFilterLink *mainlink = ctx->inputs[0]; + AVFilterLink *secondlink = ctx->inputs[1]; int ret, i, j; + s->primarywidth[1] = s->primarywidth[2] = AV_CEIL_RSHIFT(mainlink->w, desc->log2_chroma_w); + s->primarywidth[0] = s->primarywidth[3] = mainlink->w; + s->primaryheight[1] = s->primaryheight[2] = AV_CEIL_RSHIFT(mainlink->h, desc->log2_chroma_h); + s->primaryheight[0] = s->primaryheight[3] = mainlink->h; + + s->secondarywidth[1] = s->secondarywidth[2] = AV_CEIL_RSHIFT(secondlink->w, desc->log2_chroma_w); + s->secondarywidth[0] = s->secondarywidth[3] = secondlink->w; + s->secondaryheight[1] = s->secondaryheight[2] = AV_CEIL_RSHIFT(secondlink->h, desc->log2_chroma_h); + s->secondaryheight[0] = s->secondaryheight[3] = secondlink->h; + s->fs.on_event = do_convolve; ret = ff_framesync_init_dualinput(&s->fs, ctx); if (ret < 0) @@ -593,8 +802,19 @@ static av_cold int init(AVFilterContext *ctx) if (!strcmp(ctx->filter->name, "convolve")) { s->filter = complex_multiply; + s->prepare_impulse = prepare_impulse; + s->get_input = get_input; + s->get_output = get_output; + } else if (!strcmp(ctx->filter->name, "xcorrelate")) { + s->filter = complex_xcorrelate; + s->prepare_impulse = prepare_secondary; + s->get_input = get_zeropadded_input; + s->get_output = get_xoutput; } else if (!strcmp(ctx->filter->name, "deconvolve")) { s->filter = complex_divide; + s->prepare_impulse = prepare_impulse; + s->get_input = get_input; + s->get_output = get_output; } else { return AVERROR_BUG; } @@ -630,7 +850,7 @@ static const AVFilterPad convolve_inputs[] = { { .name = "main", .type = AVMEDIA_TYPE_VIDEO, - .config_props = config_input_main, + .config_props = config_input, },{ .name = "impulse", .type = AVMEDIA_TYPE_VIDEO, @@ -698,3 +918,65 @@ const AVFilter ff_vf_deconvolve = { }; #endif /* CONFIG_DECONVOLVE_FILTER */ + +#if CONFIG_XCORRELATE_FILTER + +static const AVOption xcorrelate_options[] = { + { "planes", "set planes to cross-correlate", OFFSET(planes), AV_OPT_TYPE_INT, {.i64=7}, 0, 15, FLAGS }, + { "secondary", "when to process secondary frame", OFFSET(impulse), AV_OPT_TYPE_INT, {.i64=1}, 0, 1, FLAGS, "impulse" }, + { "first", "process only first secondary frame, ignore rest", 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, FLAGS, "impulse" }, + { "all", "process all secondary frames", 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, FLAGS, "impulse" }, + { NULL }, +}; + +FRAMESYNC_DEFINE_PURE_CLASS(xcorrelate, "xcorrelate", convolve, xcorrelate_options); + +static int config_input_secondary(AVFilterLink *inlink) +{ + AVFilterContext *ctx = inlink->dst; + + if (ctx->inputs[0]->w <= ctx->inputs[1]->w || + ctx->inputs[0]->h <= ctx->inputs[1]->h) { + av_log(ctx, AV_LOG_ERROR, "Width and height of second input videos must be less than first input.\n"); + return AVERROR(EINVAL); + } + + return 0; +} + +static const AVFilterPad xcorrelate_inputs[] = { + { + .name = "primary", + .type = AVMEDIA_TYPE_VIDEO, + .config_props = config_input, + },{ + .name = "secondary", + .type = AVMEDIA_TYPE_VIDEO, + .config_props = config_input_secondary, + }, +}; + +static const AVFilterPad xcorrelate_outputs[] = { + { + .name = "default", + .type = AVMEDIA_TYPE_VIDEO, + .config_props = config_output, + }, +}; + +const AVFilter ff_vf_xcorrelate = { + .name = "xcorrelate", + .description = NULL_IF_CONFIG_SMALL("Cross-correlate first video stream with second video stream."), + .preinit = convolve_framesync_preinit, + .init = init, + .uninit = uninit, + .activate = activate, + .priv_size = sizeof(ConvolveContext), + .priv_class = &xcorrelate_class, + FILTER_INPUTS(xcorrelate_inputs), + FILTER_OUTPUTS(xcorrelate_outputs), + FILTER_PIXFMTS_ARRAY(pixel_fmts_fftfilt), + .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL | AVFILTER_FLAG_SLICE_THREADS, +}; + +#endif /* CONFIG_XCORRELATE_FILTER */