[FFmpeg-devel] [PATCH 1/2] avfilter/vf_gblur: fix divide by zero

zhilizhao quinkblack at foxmail.com
Mon Jan 6 07:11:52 EET 2020



> On Jan 4, 2020, at 9:43 PM, zhilizhao <quinkblack at foxmail.com> wrote:
> 
> 
> 
>> On Jan 4, 2020, at 9:09 PM, Paul B Mahol <onemda at gmail.com <mailto:onemda at gmail.com>> wrote:
>> 
>> On 1/4/20, zhilizhao <quinkblack at foxmail.com <mailto:quinkblack at foxmail.com>> wrote:
>>> 
>>> 
>>>> On Jan 4, 2020, at 7:04 PM, Paul B Mahol <onemda at gmail.com <mailto:onemda at gmail.com>> wrote:
>>>> 
>>>> I do not like this “fix"
>>> 
>>> What do you suggest?  It’s numerical unstable for small sigma.
>> 
>> For small sigma return frames unchanged as it currently does.
> 
> Then which valud should be selected as the threshold:
> 
> 1. A value from which it becomes unstable (it’s not fun to figure it out).
> 
> 2. A reasonable (and random) selected value like 0.25
> 

For function set_params:

static void set_params(float sigma, int steps, float *postscale, float *boundaryscale, float *nu)
{
    double dnu, lambda;

    lambda = (sigma * sigma) / (2.0 * steps);
    dnu = (1.0 + 2.0 * lambda - sqrt(1.0 + 4.0 * lambda)) / (2.0 * lambda);
    *postscale = pow(dnu / lambda, steps);
    *boundaryscale = 1.0 / (1.0 - dnu);
    *nu = (float)dnu;
}

1. Check sigma against 0 is not enough, since lambda can be 0 for non-zero sigma
2. Check lambda against 0 is not enough, we should avoid dividing by small value, like 'dnu / lambda’.

So there should have a dead-zone for sigma.

> 
>> 
>>> 
>>>> 
>>>> On 1/4/20, quinkblack at foxmail.com <quinkblack at foxmail.com> wrote:
>>>>> From: Zhao Zhili <zhilizhao at tencent.com>
>>>>> 
>>>>> ./ffmpeg -i ~/Pictures/test.jpg -vf 'gblur=sigma=0' -f null -
>>>>> ...
>>>>> src/libavfilter/vf_gblur.c:260:59: runtime error: division by zero
>>>>> src/libavfilter/vf_gblur.c:261:26: runtime error: division by zero
>>>>> ---
>>>>> libavfilter/vf_gblur.c | 12 +++++++++---
>>>>> 1 file changed, 9 insertions(+), 3 deletions(-)
>>>>> 
>>>>> diff --git a/libavfilter/vf_gblur.c b/libavfilter/vf_gblur.c
>>>>> index 2e587f6a0a..e057937fb7 100644
>>>>> --- a/libavfilter/vf_gblur.c
>>>>> +++ b/libavfilter/vf_gblur.c
>>>>> @@ -37,11 +37,17 @@
>>>>> #define OFFSET(x) offsetof(GBlurContext, x)
>>>>> #define FLAGS
>>>>> AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_RUNTIME_PARAM
>>>>> 
>>>>> +/* Consider the three-sigma rule, for minimum radius of 1 sigma should
>>>>> not
>>>>> + * be smaller than 1/3. Relax it to 0.25 if the user want to try.
>>>>> + */
>>>>> +#define SIGMA_MIN   0.25
>>>>> +#define SIGMA_MAX   1024.0
>>>>> +
>>>>> static const AVOption gblur_options[] = {
>>>>> -    { "sigma",  "set sigma",            OFFSET(sigma),
>>>>> AV_OPT_TYPE_FLOAT,
>>>>> {.dbl=0.5}, 0.0, 1024, FLAGS },
>>>>> +    { "sigma",  "set sigma",            OFFSET(sigma),
>>>>> AV_OPT_TYPE_FLOAT,
>>>>> {.dbl=0.5},  SIGMA_MIN, SIGMA_MAX, FLAGS },
>>>>>   { "steps",  "set number of steps",  OFFSET(steps),  AV_OPT_TYPE_INT,
>>>>> {.i64=1},     1,    6, FLAGS },
>>>>>   { "planes", "set planes to filter", OFFSET(planes), AV_OPT_TYPE_INT,
>>>>> {.i64=0xF},   0,  0xF, FLAGS },
>>>>> -    { "sigmaV", "set vertical sigma",   OFFSET(sigmaV),
>>>>> AV_OPT_TYPE_FLOAT,
>>>>> {.dbl=-1},   -1, 1024, FLAGS },
>>>>> +    { "sigmaV", "set vertical sigma",   OFFSET(sigmaV),
>>>>> AV_OPT_TYPE_FLOAT,
>>>>> {.dbl=-1},   -1, SIGMA_MAX, FLAGS },
>>>>>   { NULL }
>>>>> };
>>>>> 
>>>>> @@ -244,7 +250,7 @@ static int config_input(AVFilterLink *inlink)
>>>>>   if (!s->buffer)
>>>>>       return AVERROR(ENOMEM);
>>>>> 
>>>>> -    if (s->sigmaV < 0) {
>>>>> +    if (s->sigmaV < SIGMA_MIN) {
>>>>>       s->sigmaV = s->sigma;
>>>>>   }
>>>>>   ff_gblur_init(s);
>>>>> --
>>>>> 2.22.0
>>>>> 
>>>>> _______________________________________________
>>>>> 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".
>>> 
>>> 
>>> 
>>> _______________________________________________
>>> 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".
> 
> 
> 
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org <mailto:ffmpeg-devel at ffmpeg.org>
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel <https://ffmpeg.org/mailman/listinfo/ffmpeg-devel>
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org <mailto:ffmpeg-devel-request at ffmpeg.org> with subject "unsubscribe".



More information about the ffmpeg-devel mailing list