On Sat, Feb 24, 2024 at 6:34 PM Marton Balint <c...@passwd.hu> wrote:
>
>
> On Sat, 24 Feb 2024, Stone Chen wrote:
>
> > On Sat, Feb 24, 2024 at 3:56 PM Marton Balint <c...@passwd.hu> wrote:
> >
> >>
> >>
> >> On Sat, 24 Feb 2024, Stone Chen wrote:
> >>
> >> > Previously to support dynamic reconfigurations of the matrix string
> >> (e.g. 0m), the rdiv values would always be cleared to 0.f, causing the
> rdiv
> >> to be recalculated based on the new filter. This however had the side
> >> effect of always ignoring user specified rdiv values.
> >> >
> >> > Instead float user_rdiv[0] is added to ConvolutionContext which will
> >> store the user specified rdiv values. Then the original rdiv array will
> >> store either the user_rdiv or the automatically calculated 1/sum.
> >> >
> >> > This fixes trac #10294, #10867
> >>
> >> Have you tested?
> >>
> >> Thanks,
> >> Marton
> >>
> >
> >
> > Hi Marton,
> >
> > Yes I've tested that
> >
> > - the original behavior works (automatically calculate rdiv if
> user_rdiv
> > = 0)
> > - setting the rdiv value works
>
> It does not work for me even after applying your patch.
>
> Regards,
> Marton
>
Hi Marton,
Sorry about that, was it with the attached patch in the second email? I had
botched my initial patch, I've re-attached the second to this email.
If it was, I'll need to investigate further then, would you be able to send
me what you did to test?
Regards,
Stone
From d489e66c7f1ea94ef302759566ee2b22b7895b86 Mon Sep 17 00:00:00 2001
From: Stone Chen <chen.stonec...@gmail.com>
Date: Sat, 24 Feb 2024 11:08:02 -0500
Subject: [PATCH] Add float user_rdiv[4] to allow user options to apply
correctly
Previously to support dynamic reconfigurations of the matrix string (e.g. 0m), the rdiv values would always be cleared to 0.f, causing the rdiv to be recalculated based on the new filter. This however had the side effect of always ignoring user specified rdiv values.
Instead float user_rdiv[0] is added to ConvolutionContext which will store the user specified rdiv values. Then the original rdiv array will store either the user_rdiv or the automatically calculated 1/sum.
This fixes trac #10294, #10867
Signed-off-by: Stone Chen <chen.stonec...@gmail.com>
---
libavfilter/convolution.h | 3 ++-
libavfilter/vf_convolution.c | 10 +++++-----
2 files changed, 7 insertions(+), 6 deletions(-)
diff --git a/libavfilter/convolution.h b/libavfilter/convolution.h
index e44bfb5da8..ee7477ef89 100644
--- a/libavfilter/convolution.h
+++ b/libavfilter/convolution.h
@@ -34,13 +34,14 @@ typedef struct ConvolutionContext {
const AVClass *class;
char *matrix_str[4];
- float rdiv[4];
+ float user_rdiv[4];
float bias[4];
int mode[4];
float scale;
float delta;
int planes;
+ float rdiv[4];
int size[4];
int depth;
int max;
diff --git a/libavfilter/vf_convolution.c b/libavfilter/vf_convolution.c
index bf67f392f6..88b89289a9 100644
--- a/libavfilter/vf_convolution.c
+++ b/libavfilter/vf_convolution.c
@@ -40,10 +40,10 @@ static const AVOption convolution_options[] = {
{ "1m", "set matrix for 2nd plane", OFFSET(matrix_str[1]), AV_OPT_TYPE_STRING, {.str="0 0 0 0 1 0 0 0 0"}, 0, 0, FLAGS },
{ "2m", "set matrix for 3rd plane", OFFSET(matrix_str[2]), AV_OPT_TYPE_STRING, {.str="0 0 0 0 1 0 0 0 0"}, 0, 0, FLAGS },
{ "3m", "set matrix for 4th plane", OFFSET(matrix_str[3]), AV_OPT_TYPE_STRING, {.str="0 0 0 0 1 0 0 0 0"}, 0, 0, FLAGS },
- { "0rdiv", "set rdiv for 1st plane", OFFSET(rdiv[0]), AV_OPT_TYPE_FLOAT, {.dbl=0.0}, 0.0, INT_MAX, FLAGS},
- { "1rdiv", "set rdiv for 2nd plane", OFFSET(rdiv[1]), AV_OPT_TYPE_FLOAT, {.dbl=0.0}, 0.0, INT_MAX, FLAGS},
- { "2rdiv", "set rdiv for 3rd plane", OFFSET(rdiv[2]), AV_OPT_TYPE_FLOAT, {.dbl=0.0}, 0.0, INT_MAX, FLAGS},
- { "3rdiv", "set rdiv for 4th plane", OFFSET(rdiv[3]), AV_OPT_TYPE_FLOAT, {.dbl=0.0}, 0.0, INT_MAX, FLAGS},
+ { "0rdiv", "set rdiv for 1st plane", OFFSET(user_rdiv[0]), AV_OPT_TYPE_FLOAT, {.dbl=0.0}, 0.0, INT_MAX, FLAGS},
+ { "1rdiv", "set rdiv for 2nd plane", OFFSET(user_rdiv[1]), AV_OPT_TYPE_FLOAT, {.dbl=0.0}, 0.0, INT_MAX, FLAGS},
+ { "2rdiv", "set rdiv for 3rd plane", OFFSET(user_rdiv[2]), AV_OPT_TYPE_FLOAT, {.dbl=0.0}, 0.0, INT_MAX, FLAGS},
+ { "3rdiv", "set rdiv for 4th plane", OFFSET(user_rdiv[3]), AV_OPT_TYPE_FLOAT, {.dbl=0.0}, 0.0, INT_MAX, FLAGS},
{ "0bias", "set bias for 1st plane", OFFSET(bias[0]), AV_OPT_TYPE_FLOAT, {.dbl=0.0}, 0.0, INT_MAX, FLAGS},
{ "1bias", "set bias for 2nd plane", OFFSET(bias[1]), AV_OPT_TYPE_FLOAT, {.dbl=0.0}, 0.0, INT_MAX, FLAGS},
{ "2bias", "set bias for 3rd plane", OFFSET(bias[2]), AV_OPT_TYPE_FLOAT, {.dbl=0.0}, 0.0, INT_MAX, FLAGS},
@@ -674,7 +674,7 @@ static int param_init(AVFilterContext *ctx)
p = orig = av_strdup(s->matrix_str[i]);
if (p) {
s->matrix_length[i] = 0;
- s->rdiv[i] = 0.f;
+ s->rdiv[i] = s->user_rdiv[i];
sum = 0.f;
while (s->matrix_length[i] < 49) {
--
2.43.2
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
