ffDiaporama Forum / Dead / stuck pixel interpolation

Re: Dead / stuck pixel interpolation

dummy@example.com (domledom) — Sat, 04 Jan 2014 07:15:00 +0000

The problem is that this filter is not OpenSource and so I can not integrate it into ffDiaporama

Moreover, it should not be incorporated into ffDiaporama but rather into ffmpeg.
Maybe you can contact them for they integrate it

Re: Dead / stuck pixel interpolation

dummy@example.com (palindrom) — Thu, 02 Jan 2014 09:51:08 +0000

On the other hand, it would be silly to reinvent the wheel. The filter is out there (just found it):

http://www.playtool.com/pages/pixelpatc … tcher.html

Re: Dead / stuck pixel interpolation

dummy@example.com (palindrom) — Thu, 12 Dec 2013 08:50:18 +0000

I'm working on PhD right now, seems like "new cool feature" will have to wait a while It was an indication anyway.
Cheers,
Marcin

Re: Dead / stuck pixel interpolation

dummy@example.com (domledom) — Wed, 11 Dec 2013 16:07:09 +0000

At the moment I'm working on another major innovation and can not work on that now.
But, if you propose me some code, I would consider to add them to the next version.

Dead / stuck pixel interpolation

dummy@example.com (palindrom) — Wed, 11 Dec 2013 11:06:31 +0000

Hi,
I've been using ffDiaporama for some time now, mostly using compact camera and still images. Now I have Pentax k-r and discovered a whole new world of DSLR hot-pixel-stuffed video. I believe many filmmakers will go to DSLR as it is cheap option to get one plus manual bright prime lens. DSLRs in general don't map dead pixels in video mode and it really sucks.
---
Have you considered adding simple tool to manually select deal pixel (small region actually) on 1:1 or magnified view, that would interpolate its value to average of neighbors? It would be quite unique feature and I think many would praise you for it. Even commercial software don't offer such tool even though it's not complicated operation.
---
Pro version I: automated tool, that uses dark frames footage. Camera mode usually tends to boost ISO when there's not enough light. Making short clip with lens cap on (separately for every resolution, after warming up sensor) would give few hundred expositions with random noise and high/low signal hot/dead pixels, easily found by averaging at least 10 subsequent frames, this would reduce truly random noise by factor of sqrt(n), where n is number of averaged samples. Spotting pixels is then easy:
Dead(x,y) for (I(x,y) - average_across_sensor) < treshold_dead
Hot(x,y) for (I(x,y) - average_across_sensor) > treshold_hot
---
Pro version II: detect hot/dead pixels/regions based on difference in value amplitude of surrounding pixels. It might be overkill though.
---
What do you think about that?
Regards,
Marcin Szczurowski