avfilter: add premultiply filter
authorPaul B Mahol <onemda@gmail.com>
Mon, 28 Nov 2016 16:28:59 +0000 (17:28 +0100)
committerPaul B Mahol <onemda@gmail.com>
Thu, 1 Dec 2016 21:12:21 +0000 (22:12 +0100)
Signed-off-by: Paul B Mahol <onemda@gmail.com>
Changelog
doc/filters.texi
libavfilter/Makefile
libavfilter/allfilters.c
libavfilter/version.h
libavfilter/vf_premultiply.c [new file with mode: 0644]

index 2778304..71c9273 100644 (file)
--- a/Changelog
+++ b/Changelog
@@ -5,6 +5,7 @@ version <next>:
 - CrystalHD decoder moved to new decode API
 - add internal ebur128 library, remove external libebur128 dependency
 - Pro-MPEG CoP #3-R2 FEC protocol
+- premultiply video filter
 
 version 3.2:
 - libopenmpt demuxer
index 4a4aeca..3ae3c46 100644 (file)
@@ -10983,6 +10983,12 @@ Set medium thresholding (good results, default).
 @end table
 @end table
 
+@section premultiply
+Apply alpha premultiply effect to input video stream using first plane
+of second stream as alpha.
+
+Both streams must have same dimensions and same pixel format.
+
 @section prewitt
 Apply prewitt operator to input video stream.
 
index cdddb1b..cb614c9 100644 (file)
@@ -240,6 +240,7 @@ OBJS-$(CONFIG_PHASE_FILTER)                  += vf_phase.o
 OBJS-$(CONFIG_PIXDESCTEST_FILTER)            += vf_pixdesctest.o
 OBJS-$(CONFIG_PP_FILTER)                     += vf_pp.o
 OBJS-$(CONFIG_PP7_FILTER)                    += vf_pp7.o
+OBJS-$(CONFIG_PREMULTIPLY_FILTER)            += vf_premultiply.o framesync.o
 OBJS-$(CONFIG_PREWITT_FILTER)                += vf_convolution.o
 OBJS-$(CONFIG_PSNR_FILTER)                   += vf_psnr.o dualinput.o framesync.o
 OBJS-$(CONFIG_PULLUP_FILTER)                 += vf_pullup.o
index 82a65ee..2c37818 100644 (file)
@@ -256,6 +256,7 @@ void avfilter_register_all(void)
     REGISTER_FILTER(PIXDESCTEST,    pixdesctest,    vf);
     REGISTER_FILTER(PP,             pp,             vf);
     REGISTER_FILTER(PP7,            pp7,            vf);
+    REGISTER_FILTER(PREMULTIPLY,    premultiply,    vf);
     REGISTER_FILTER(PREWITT,        prewitt,        vf);
     REGISTER_FILTER(PSNR,           psnr,           vf);
     REGISTER_FILTER(PULLUP,         pullup,         vf);
index b8c9b81..e3bd8d0 100644 (file)
@@ -30,7 +30,7 @@
 #include "libavutil/version.h"
 
 #define LIBAVFILTER_VERSION_MAJOR   6
-#define LIBAVFILTER_VERSION_MINOR  67
+#define LIBAVFILTER_VERSION_MINOR  68
 #define LIBAVFILTER_VERSION_MICRO 100
 
 #define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \
diff --git a/libavfilter/vf_premultiply.c b/libavfilter/vf_premultiply.c
new file mode 100644 (file)
index 0000000..7d276df
--- /dev/null
@@ -0,0 +1,409 @@
+/*
+ * Copyright (c) 2016 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 "framesync.h"
+#include "internal.h"
+#include "video.h"
+
+typedef struct PreMultiplyContext {
+    const AVClass *class;
+    int width[4], height[4];
+    int nb_planes;
+    int planes;
+    int half, depth, offset;
+    FFFrameSync fs;
+
+    void (*premultiply[4])(const uint8_t *msrc, const uint8_t *asrc,
+                           uint8_t *dst,
+                           ptrdiff_t mlinesize, ptrdiff_t alinesize,
+                           ptrdiff_t dlinesize,
+                           int w, int h,
+                           int half, int shift, int offset);
+} PreMultiplyContext;
+
+#define OFFSET(x) offsetof(PreMultiplyContext, x)
+#define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
+
+static const AVOption premultiply_options[] = {
+    { NULL }
+};
+
+AVFILTER_DEFINE_CLASS(premultiply);
+
+static int query_formats(AVFilterContext *ctx)
+{
+    static const enum AVPixelFormat pix_fmts[] = {
+        AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUVJ444P,
+        AV_PIX_FMT_YUV444P9, AV_PIX_FMT_YUV444P10,
+        AV_PIX_FMT_YUV444P12, AV_PIX_FMT_YUV444P14,
+        AV_PIX_FMT_YUV444P16,
+        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_GRAY8, AV_PIX_FMT_GRAY10, AV_PIX_FMT_GRAY12, AV_PIX_FMT_GRAY16,
+        AV_PIX_FMT_NONE
+    };
+
+    return ff_set_common_formats(ctx, ff_make_format_list(pix_fmts));
+}
+
+static void premultiply8(const uint8_t *msrc, const uint8_t *asrc,
+                         uint8_t *dst,
+                         ptrdiff_t mlinesize, ptrdiff_t alinesize,
+                         ptrdiff_t dlinesize,
+                         int w, int h,
+                         int half, int shift, int offset)
+{
+    int x, y;
+
+    for (y = 0; y < h; y++) {
+        for (x = 0; x < w; x++) {
+            dst[x] = ((msrc[x] * (((asrc[x] >> 1) & 1) + asrc[x])) + 128) >> 8;
+        }
+
+        dst  += dlinesize;
+        msrc += mlinesize;
+        asrc += alinesize;
+    }
+}
+
+static void premultiply8yuv(const uint8_t *msrc, const uint8_t *asrc,
+                            uint8_t *dst,
+                            ptrdiff_t mlinesize, ptrdiff_t alinesize,
+                            ptrdiff_t dlinesize,
+                            int w, int h,
+                            int half, int shift, int offset)
+{
+    int x, y;
+
+    for (y = 0; y < h; y++) {
+        for (x = 0; x < w; x++) {
+            dst[x] = ((((msrc[x] - 128) * (((asrc[x] >> 1) & 1) + asrc[x]))) >> 8) + 128;;
+        }
+
+        dst  += dlinesize;
+        msrc += mlinesize;
+        asrc += alinesize;
+    }
+}
+
+static void premultiply8offset(const uint8_t *msrc, const uint8_t *asrc,
+                               uint8_t *dst,
+                               ptrdiff_t mlinesize, ptrdiff_t alinesize,
+                               ptrdiff_t dlinesize,
+                               int w, int h,
+                               int half, int shift, int offset)
+{
+    int x, y;
+
+    for (y = 0; y < h; y++) {
+        for (x = 0; x < w; x++) {
+            dst[x] = ((((msrc[x] - offset) * (((asrc[x] >> 1) & 1) + asrc[x])) + 128) >> 8) + offset;
+        }
+
+        dst  += dlinesize;
+        msrc += mlinesize;
+        asrc += alinesize;
+    }
+}
+
+static void premultiply16(const uint8_t *mmsrc, const uint8_t *aasrc,
+                          uint8_t *ddst,
+                          ptrdiff_t mlinesize, ptrdiff_t alinesize,
+                          ptrdiff_t dlinesize,
+                          int w, int h,
+                          int half, int shift, int offset)
+{
+    const uint16_t *msrc = (const uint16_t *)mmsrc;
+    const uint16_t *asrc = (const uint16_t *)aasrc;
+    uint16_t *dst = (uint16_t *)ddst;
+    int x, y;
+
+    for (y = 0; y < h; y++) {
+        for (x = 0; x < w; x++) {
+            dst[x] = ((msrc[x] * (((asrc[x] >> 1) & 1) + asrc[x])) + half) >> shift;
+        }
+
+        dst  += dlinesize / 2;
+        msrc += mlinesize / 2;
+        asrc += alinesize / 2;
+    }
+}
+
+static void premultiply16yuv(const uint8_t *mmsrc, const uint8_t *aasrc,
+                             uint8_t *ddst,
+                             ptrdiff_t mlinesize, ptrdiff_t alinesize,
+                             ptrdiff_t dlinesize,
+                             int w, int h,
+                             int half, int shift, int offset)
+{
+    const uint16_t *msrc = (const uint16_t *)mmsrc;
+    const uint16_t *asrc = (const uint16_t *)aasrc;
+    uint16_t *dst = (uint16_t *)ddst;
+    int x, y;
+
+    for (y = 0; y < h; y++) {
+        for (x = 0; x < w; x++) {
+            dst[x] = ((((msrc[x] - half) * (((asrc[x] >> 1) & 1) + asrc[x]))) >> shift) + half;;
+        }
+
+        dst  += dlinesize / 2;
+        msrc += mlinesize / 2;
+        asrc += alinesize / 2;
+    }
+}
+
+static void premultiply16offset(const uint8_t *mmsrc, const uint8_t *aasrc,
+                                uint8_t *ddst,
+                                ptrdiff_t mlinesize, ptrdiff_t alinesize,
+                                ptrdiff_t dlinesize,
+                                int w, int h,
+                                int half, int shift, int offset)
+{
+    const uint16_t *msrc = (const uint16_t *)mmsrc;
+    const uint16_t *asrc = (const uint16_t *)aasrc;
+    uint16_t *dst = (uint16_t *)ddst;
+    int x, y;
+
+    for (y = 0; y < h; y++) {
+        for (x = 0; x < w; x++) {
+            dst[x] = ((((msrc[x] - offset) * (((asrc[x] >> 1) & 1) + asrc[x])) + half) >> shift) + offset;
+        }
+
+        dst  += dlinesize / 2;
+        msrc += mlinesize / 2;
+        asrc += alinesize / 2;
+    }
+}
+
+static int process_frame(FFFrameSync *fs)
+{
+    AVFilterContext *ctx = fs->parent;
+    PreMultiplyContext *s = fs->opaque;
+    AVFilterLink *outlink = ctx->outputs[0];
+    AVFrame *out, *base, *alpha;
+    int ret;
+
+    if ((ret = ff_framesync_get_frame(&s->fs, 0, &base,  0)) < 0 ||
+        (ret = ff_framesync_get_frame(&s->fs, 1, &alpha, 0)) < 0)
+        return ret;
+
+    if (ctx->is_disabled) {
+        out = av_frame_clone(base);
+        if (!out)
+            return AVERROR(ENOMEM);
+    } else {
+        int p, full, limited;
+
+        out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
+        if (!out)
+            return AVERROR(ENOMEM);
+        av_frame_copy_props(out, base);
+
+        full = base->color_range == AVCOL_RANGE_JPEG;
+        limited = base->color_range == AVCOL_RANGE_MPEG;
+
+        switch (outlink->format) {
+        case AV_PIX_FMT_YUV444P:
+            s->premultiply[0] = full ? premultiply8 : premultiply8offset;
+            s->premultiply[1] = premultiply8yuv;
+            s->premultiply[2] = premultiply8yuv;
+            break;
+        case AV_PIX_FMT_YUVJ444P:
+            s->premultiply[0] = premultiply8;
+            s->premultiply[1] = premultiply8yuv;
+            s->premultiply[2] = premultiply8yuv;
+            break;
+        case AV_PIX_FMT_GBRP:
+            s->premultiply[0] = limited ? premultiply8offset : premultiply8;
+            s->premultiply[1] = limited ? premultiply8offset : premultiply8;
+            s->premultiply[2] = limited ? premultiply8offset : premultiply8;
+            break;
+        case AV_PIX_FMT_YUV444P9:
+        case AV_PIX_FMT_YUV444P10:
+        case AV_PIX_FMT_YUV444P12:
+        case AV_PIX_FMT_YUV444P14:
+        case AV_PIX_FMT_YUV444P16:
+            s->premultiply[0] = full ? premultiply16 : premultiply16offset;
+            s->premultiply[1] = premultiply16yuv;
+            s->premultiply[2] = premultiply16yuv;
+            break;
+        case AV_PIX_FMT_GBRP9:
+        case AV_PIX_FMT_GBRP10:
+        case AV_PIX_FMT_GBRP12:
+        case AV_PIX_FMT_GBRP14:
+        case AV_PIX_FMT_GBRP16:
+            s->premultiply[0] = limited ? premultiply16offset : premultiply16;
+            s->premultiply[1] = limited ? premultiply16offset : premultiply16;
+            s->premultiply[2] = limited ? premultiply16offset : premultiply16;
+            break;
+        case AV_PIX_FMT_GRAY8:
+            s->premultiply[0] = limited ? premultiply8offset : premultiply8;
+            break;
+        case AV_PIX_FMT_GRAY10:
+        case AV_PIX_FMT_GRAY12:
+        case AV_PIX_FMT_GRAY16:
+            s->premultiply[0] = limited ? premultiply16offset : premultiply16;
+            break;
+        }
+
+        for (p = 0; p < s->nb_planes; p++) {
+            s->premultiply[p](base->data[p], alpha->data[0],
+                              out->data[p],
+                              base->linesize[p], alpha->linesize[0],
+                              out->linesize[p],
+                              s->width[p], s->height[p],
+                              s->half, s->depth, s->offset);
+        }
+    }
+    out->pts = av_rescale_q(base->pts, s->fs.time_base, outlink->time_base);
+
+    return ff_filter_frame(outlink, out);
+}
+
+static int config_input(AVFilterLink *inlink)
+{
+    AVFilterContext *ctx = inlink->dst;
+    PreMultiplyContext *s = ctx->priv;
+    const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
+    int vsub, hsub;
+
+    s->nb_planes = av_pix_fmt_count_planes(inlink->format);
+
+    hsub = desc->log2_chroma_w;
+    vsub = desc->log2_chroma_h;
+    s->height[1] = s->height[2] = AV_CEIL_RSHIFT(inlink->h, vsub);
+    s->height[0] = s->height[3] = inlink->h;
+    s->width[1]  = s->width[2]  = AV_CEIL_RSHIFT(inlink->w, hsub);
+    s->width[0]  = s->width[3]  = inlink->w;
+
+    s->depth = desc->comp[0].depth;
+    s->half = (1 << s->depth) / 2;
+    s->offset = 16 << (s->depth - 8);
+
+    return 0;
+}
+
+static int config_output(AVFilterLink *outlink)
+{
+    AVFilterContext *ctx = outlink->src;
+    PreMultiplyContext *s = ctx->priv;
+    AVFilterLink *base = ctx->inputs[0];
+    AVFilterLink *alpha = ctx->inputs[1];
+    FFFrameSyncIn *in;
+    int ret;
+
+    if (base->format != alpha->format) {
+        av_log(ctx, AV_LOG_ERROR, "inputs must be of same pixel format\n");
+        return AVERROR(EINVAL);
+    }
+    if (base->w                       != alpha->w ||
+        base->h                       != alpha->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) ",
+               ctx->input_pads[0].name, base->w, base->h,
+               ctx->input_pads[1].name, alpha->w, alpha->h);
+        return AVERROR(EINVAL);
+    }
+
+    outlink->w = base->w;
+    outlink->h = base->h;
+    outlink->time_base = base->time_base;
+    outlink->sample_aspect_ratio = base->sample_aspect_ratio;
+    outlink->frame_rate = base->frame_rate;
+
+    if ((ret = ff_framesync_init(&s->fs, ctx, 2)) < 0)
+        return ret;
+
+    in = s->fs.in;
+    in[0].time_base = base->time_base;
+    in[1].time_base = alpha->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;
+    s->fs.opaque   = s;
+    s->fs.on_event = process_frame;
+
+    return ff_framesync_configure(&s->fs);
+}
+
+static int filter_frame(AVFilterLink *inlink, AVFrame *buf)
+{
+    PreMultiplyContext *s = inlink->dst->priv;
+    return ff_framesync_filter_frame(&s->fs, inlink, buf);
+}
+
+static int request_frame(AVFilterLink *outlink)
+{
+    PreMultiplyContext *s = outlink->src->priv;
+    return ff_framesync_request_frame(&s->fs, outlink);
+}
+
+static av_cold void uninit(AVFilterContext *ctx)
+{
+    PreMultiplyContext *s = ctx->priv;
+
+    ff_framesync_uninit(&s->fs);
+}
+
+static const AVFilterPad premultiply_inputs[] = {
+    {
+        .name         = "main",
+        .type         = AVMEDIA_TYPE_VIDEO,
+        .filter_frame = filter_frame,
+        .config_props = config_input,
+    },
+    {
+        .name         = "alpha",
+        .type         = AVMEDIA_TYPE_VIDEO,
+        .filter_frame = filter_frame,
+    },
+    { NULL }
+};
+
+static const AVFilterPad premultiply_outputs[] = {
+    {
+        .name          = "default",
+        .type          = AVMEDIA_TYPE_VIDEO,
+        .config_props  = config_output,
+        .request_frame = request_frame,
+    },
+    { NULL }
+};
+
+AVFilter ff_vf_premultiply = {
+    .name          = "premultiply",
+    .description   = NULL_IF_CONFIG_SMALL("PreMultiply first stream with first plane of second stream."),
+    .priv_size     = sizeof(PreMultiplyContext),
+    .uninit        = uninit,
+    .query_formats = query_formats,
+    .inputs        = premultiply_inputs,
+    .outputs       = premultiply_outputs,
+    .priv_class    = &premultiply_class,
+    .flags         = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL,
+};