Magenta edge tinging at high ISO and a possible solution...

Started Aug 17, 2009 | Discussions thread
GordonBGood Veteran Member • Posts: 6,308
Re: PROGRESS: Version 0.8.0.0 has one more tiny fix...

GordonBGood wrote:

dlacoutu wrote:

Well, I jumped on your successive versions to give them a try, and they work quite well... Magenta noise is now better balanced (and manageable) across the picture, and 6400 ISO can be used now.

But (there is always a "but"), there is still a tiny magenta ridge near the bottom edge of my pics.

Now, there are a limited number of explanations of why this magenta ridge is still left over that I can think of off hand, as follows:

1) That the magenta ridge is too narrow to be caught by the blurring that I do and the correction can't follow it because the correction "pen" is too broad.

2) If the magenta ridge is less than 80 photosites from the bottom edge of the image and thus in the linear correction range. I don't think that this is true as I recall, as I don't have your sample here, but it's possible.

3) That the ridge in the image doesn't have a corresponding ridge in the masked black border area used for correction and thus the correction algorithm never sees the ridge.

4) That the magenta ridge is within the 128 pixel zone that I use to extrapolate out to the edge of the sensor and thus throws out the extrapolation.

I should be able to check for these cases quite quickly when I go in to my project computer in a few hours. If I find it to be due to one of these reasons, I should be able to modify the algorithm slightly to accommodate it.

I did check for these and most of these don't apply, as 1) the magenta ridge is wide enough to be caught by my processing, 2) and 4) it's far enough from the bottom to still be corrected, which leaves 3) that the ridge doesn't have a corresponding ridge in the black masked border at the right side, I therefore don't detect it, and thus don't correct it. I think this is another case of the same sort of wide pattern noise banding problem that godfrog sees with his sensor vertically except that yours is horizontal and a little bit narrower. If I am correct that the black offset correction algorithm causes this and removes the evidence from masked black border photosites, I won't be able to automatically detect or correct these just as for Vertical Pattern Noise (VPN) for the K10D.

I believe that these sorts of banding problems are camera sample specific, and not all samples will have it.

At least this correction program gets rid of most of the border tingeing automatically. To get rid of these horizontal/vertical banding that can't be detected in the masked black borders, the only solution would be to build a model of a specific camera as it varies with temperature to be applied automatically, or to correct it manually with an offset that is applied over a range of columns or rows, which would be a separate project to this one. This is a similar problem to the "thin green vertical line" problem that occurs for some samples of the K-7. I may start such a project in a month or two one this one is up and stable, but I have limited time to commit to it.

Regards, GordonBGood

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow