[FFmpeg-user] Documentation of chromanr filter

[Paul B Mahol]

Please send patch, otherwise it have little sense to discuss this.

On 7/11/20, Michael Koch <astroelectronic at t-online.de> wrote:
> Hello,
>
> The documentation of the new chromanr filter, especially this sentence
> for the "thres" option is hard to understand:
>
> "Sum of absolute difference of U and V pixel components or current pixel
> and neighbour pixels lower than this threshold will be used in averaging."
>
> First of all I think "or" is a typo and should be "of". But even with
> this change the sentence is still hard to understand.
> The sum of absolute differences of what? Sum of absolute differences of
> the U and V pixel components? That makes no sense.
>
> This is how I think it might work:
> Foreach neighbour pixel
> {
>    A = absolute difference of U components of current pixel and
> neighbour pixel
>    B = absolute difference of V components of current pixel and
> neighbour pixel
>    if (A+B < threshold) then
>      use this neighbour pixel for averaging
> }
>
> If the above is correct, I suggest to change the sentence as follows:
>
> "The algorithm calculates the absolute difference of the U components of
> the current pixel and a neighbour pixel. The same is also calculated for
> the V components. A neighbour pixel is used for averaging, if the sum of
> both absolute differences is lower than the threshold."
>
> Also the meaning of  "stepw" and "steph" options is unclear:
> -- Do these options refer to the pixels? If stepw=2 and steph=2 then
> only a quarter of the pixels is analyzed, and the same U and V
> components are used for 4 pixels?
> -- Or do they refer to the selection of the neighbour pixels? If stepw=2
> and steph=2 then only a quarter of the neighbour pixels is analyzed?
>
> Michael
> _______________________________________________
> ffmpeg-user mailing list
> ffmpeg-user at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-user
>
> To unsubscribe, visit link above, or email
> ffmpeg-user-request at ffmpeg.org with subject "unsubscribe".