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

Started Aug 17, 2009 | Discussions thread
GordonBGood Veteran Member • Posts: 6,308
Re: Magenta edge tinging at high ISO and a possible solution...

godfrog wrote:

Just took a look at what ISO 6400 looks like on my GX-20, shooting a lens cap in RAW, developed in Lightroom. Images are posted below, the the resized shots have +4 exposure (aka this is ISO 102400) and +150 brightness, to better show what the pattern looks like. The 100% crop shot does not have the +150 brightness, only +4 exposure.

Looks like correcting for this problem might be difficult on my camera, as the tinging varies from shot to shot (look at the top left corner, it has lots of green in the second shot). A greater issue is the hard vertical edge where the noise changes intensity. Ive included a real shot to show the effect this has, this shot is ISO 4500.

As you say, it looks like your camera has two problems, and only one of these will this current correction program eliminate, as follows:

1) Your camera has the hard vertical edge at the left side in landscape mode which is likely due to some small defect in your sensor, most likely a hot/stuck photosite in the masked black photosite strip along the bottom of the sensor (landscape) used for black offset determination that is causing an improper offset for this whole strip. You could look for a hot pixel in that area using a program that shows the full area of the sensor such as Faststone ( http://www.faststone.org ) at high zooms. As to correcting for it, the current program does not do it as it assumes that the changes across the sensor will be gradual and not abrupt and would eliminate any hottish photosites before it does the trend line calculation anyway. Full correction for this would require that one knows exactly how the camera does black offset compensation, and although I have made some guesses in this thread, no one but Pentax/Samsung will know for sure. However, your camera may offer some clues, especially if there are some hot photosite(s) in the mentioned black strip, so if you would care to send me a link to download one or both of the raw files for the first two examples, I can take a look. If I can come close to guessing how the camera algorithms work, I can improve the algorithm as to correcting the things Pentax/Samsung miss and possibly add something that will automatically take care of this hard band along with the second problem already described.

Now looking harder at the hard edge from dark to bright going from right to left at the left edge of your samples, one might think that the error is that the left edge is too bright, but in actual fact the edge might be caused by the next rightmost band being too dark, which is what I would expect from hot/stuck photosites in the next rightmost band in the bottom border black band. Downloading your 100% crop, I already see that the bands are about 335 pixels wide which mean the image is divided into 14 vertical correction bands, but I would like to confirm the exact pattern from the raw data. Based on what I see here, I should be able to take care of this problem just as I do for the second problem already described as long as I can establish the exact strip boundaries and a reasonable guess as to the Pentax correction algorithm, much as I was able to reduce the K10D/GX10's Vertical Pattern Noise (VPN) due to hot pixels in this masked strip for firmware version 1.0.

Question at large: do some samples of the K20D also sometimes show these vertical strips when push processed as described here? Running your camera in Live View mode and thus heating the sensor is likely to make the problem worse, as it would increase the presence of hot photosites, as would longer exposure times up to the point where Dark Frame Subtraction (DFS) cuts in. I see that these images were taken at a one second exposure but the MakerNotes tag has been stripped from the EXIF information and I can't see the sensor temperature (PhotoMe - http://www.photome.de ).

2) The fringing toward the top and bottom edges in landscape orientation is exactly what this correction program is designed to correct and the fact that the fringe colour varies from shot to shot doesn't concern me in the least as the correction is designed to be self adjusting due to being based on the trends established from the masked black offset determination photosites for each colour channel on the right and bottom sides of the sensor (landscape orientation).

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