[FFmpeg-devel] [PATCH 1/3] lavf/f_select: support scenecut with more pixel formats

[Limin Wang]

On Wed, Jul 17, 2019 at 12:58:04AM +0200, Marton Balint wrote:
> 
> 
> On Tue, 16 Jul 2019, lance.lmwang at gmail.com wrote:
> 
> >From: Limin Wang <lance.lmwang at gmail.com>
> >
> >This patch haven't make other pixel format usable yet to make sure the test
> >result is same with rgb format.
> >
> >Signed-off-by: Limin Wang <lance.lmwang at gmail.com>
> >---
> >libavfilter/f_select.c | 34 ++++++++++++++++++++++++++++++----
> >1 file changed, 30 insertions(+), 4 deletions(-)
> >
> >diff --git a/libavfilter/f_select.c b/libavfilter/f_select.c
> >index 1132375..eed8df3 100644
> >--- a/libavfilter/f_select.c
> >+++ b/libavfilter/f_select.c
> >@@ -28,6 +28,8 @@
> >#include "libavutil/fifo.h"
> >#include "libavutil/internal.h"
> >#include "libavutil/opt.h"
> >+#include "libavutil/imgutils.h"
> >+#include "libavutil/pixdesc.h"
> >#include "avfilter.h"
> >#include "audio.h"
> >#include "formats.h"
> >@@ -144,6 +146,10 @@ typedef struct SelectContext {
> >    char *expr_str;
> >    AVExpr *expr;
> >    double var_values[VAR_VARS_NB];
> >+    int bitdepth;
> >+    int nb_planes;
> >+    ptrdiff_t width[4];
> >+    ptrdiff_t height[4];
> >    int do_scene_detect;            ///< 1 if the expression requires scene detection variables, 0 otherwise
> >    ff_scene_sad_fn sad;            ///< Sum of the absolute difference function (scene detect only)
> >    double prev_mafd;               ///< previous MAFD                           (scene detect only)
> >@@ -202,6 +208,17 @@ static av_cold int init(AVFilterContext *ctx)
> >static int config_input(AVFilterLink *inlink)
> >{
> >    SelectContext *select = inlink->dst->priv;
> >+    const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
> >+
> >+    select->bitdepth = desc->comp[0].depth;
> >+    select->nb_planes = av_pix_fmt_count_planes(inlink->format);
> >+    for (int plane = 0; plane < select->nb_planes; plane++) {
> >+        ptrdiff_t line_size = av_image_get_linesize(inlink->format, inlink->w, plane);
> >+        int vsub = desc->log2_chroma_h;
> >+
> >+        select->width[plane] = line_size >> (select->bitdepth > 8);
> >+        select->height[plane] = plane == 1 || plane == 2 ?  AV_CEIL_RSHIFT(inlink->h, vsub) : inlink->h;
> >+    }
> >
> >    select->var_values[VAR_N]          = 0.0;
> >    select->var_values[VAR_SELECTED_N] = 0.0;
> >@@ -242,7 +259,7 @@ static int config_input(AVFilterLink *inlink)
> >        inlink->type == AVMEDIA_TYPE_AUDIO ? inlink->sample_rate : NAN;
> >
> >    if (CONFIG_SELECT_FILTER && select->do_scene_detect) {
> >-        select->sad = ff_scene_sad_get_fn(8);
> >+        select->sad = ff_scene_sad_get_fn(select->bitdepth == 8 ? 8 : 16);
> >        if (!select->sad)
> >            return AVERROR(EINVAL);
> >    }
> >@@ -258,12 +275,21 @@ static double get_scene_score(AVFilterContext *ctx, AVFrame *frame)
> >    if (prev_picref &&
> >        frame->height == prev_picref->height &&
> >        frame->width  == prev_picref->width) {
> >-        uint64_t sad;
> >+        uint64_t sad = 0;
> >        double mafd, diff;
> >+        int count = 0;
> >+
> >+        for (int plane = 0; plane < select->nb_planes; plane++) {
> >+            uint64_t plane_sad;
> >+            select->sad(prev_picref->data[plane], prev_picref->linesize[plane],
> >+                    frame->data[plane], frame->linesize[plane],
> >+                    select->width[plane], select->height[plane], &plane_sad);
> >+            sad += plane_sad;
> >+            count += select->width[plane] * select->height[plane];
> >+        }
> >
> >-        select->sad(prev_picref->data[0], prev_picref->linesize[0], frame->data[0], frame->linesize[0], frame->width * 3, frame->height, &sad);
> >        emms_c();
> >-        mafd = (double)sad / (frame->width * 3 * frame->height);
> >+        mafd = (double)sad / count;
> 
> mafd was in [0..255] before you added >8 bitdepth support. For >8
> bit you have to normalize it to this range becuase the metric should
> be bitdepth invariant.
Yes, I think 100 should be change to 1ULL << bitdepth. However if change
it, it'll make the threshold is very different with old way. So I
haven't clear how to touched it to keep constant with the old. 

I'll try get one HDR 10bit sample to test the 10bit, I guess my current 
10bit sample isn't real 10bit.


> 
> Regards,
> Marton
> 
> >        diff = fabs(mafd - select->prev_mafd);
> >        ret  = av_clipf(FFMIN(mafd, diff) / 100., 0, 1);
> >        select->prev_mafd = mafd;
> >-- 
> >2.6.4
> >
> >_______________________________________________
> >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".
> _______________________________________________
> 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".