Black Shading - rescue for footage with black shading issues

[Alex Parkinson]

Hi all,

Is there a way to fix footage with pretty severe black shading issues? (lots of hot pixels)

The footage in question was shot at 1 sec exposures and no black shade performed.

Static camera.

Thanks in advance.

Alex

[Dan Hudgins]

In my new de-Bayer program I have an option to do both FPN subtraction and bad pixel interpolation on RGB files, such as TIF or DPX.

I have not tried it yet with RED footage, but it might work, or if there is a bug it might work after that's fixed.

I use it all the time on True RAW sensor data that has not had ANY black calibration in the camera.

For doing FPN and bad pixel automatically, you need to record some black, like 15 seconds showing the same issue.

If you can't do that, then you can take a frame and mark the bad pixels in red and null out the remainder of the image, its best to like convert the sample frame to gray scale and set white at 50%, then you can mark the hot pixels in red and adjust the black level so every thing else is black, my program can then read that and make the bad pixel list to use for interpolating over the bad pixels. You cannot do the FPN manually probably, but the hot pixels the ones that are very bright can be marked by hand from a sample frame.

Those commands are mostly for used with cameras that record uncompressed so the dark calibration can be done before or after shooting. I added the RGB to RGB masking in case people using a compressed camera got a bad pixel while shooting after they did dark calibration since for wavelet compressed cameras the dark calibration needs to be done ONLY before recording for the in camera masking to work, but it can be done after in a program like mine if its possible to record the same FPN on a black frame set, or at least manually mark the very bad hot pixels.

What you would do is make TIF from REDCINE-X, process them in DANCINEC.EXE (tm) then get a result TIF or DPX set that are cleaned up, and then convert those into whatever you need.

If you have heavy 'shot noise' that changes frame to frame, that is another problem, unless its restricted to just some pixels that can be removed and interpolated over. Totally random 'shot noise' is more like film dust, so requires a different algorithm, maybe using film dust-busting software would help with that issue?

CMOS sensor noise comes in three forms:

1) Fixed, like the vertical lines that show up on CMOS sensors at high ISO or hard grading, those can be subtracted by the FPN masking from the average of many dark frames.

2) Random, that is due to heat and cosmic rays etc. it is small like film grain and changes frame to frame without a pattern. This is removed with active peaking to reduce sharpening areas of the image that contain only low contrast detail mixed with grain noise.

3) Moving, this is a type of digital crosstalk and is like interference between two TV stations in the analog days, you see flashing bars and such mostly horizontal that come and go from frame to frame. The best way to reduce these is using temporal noise resuction.

Hot pixels are not really noise, they are constant in all frames, and can easily be interpolated over since they don't move around.