CCDBand-Aid repairs vertical bars in Kodak KAI-11000M images

[bob_franke]

Hello everyone,

With version 1.2, CCDBand-Aid now repairs the vertical pattern in PixInsight files. The program assumes the default PI settings. That is, 32-bit floating-point FITS images with pixel values from 0.0 to 1.0.

CCDBand-Aid determines the phase and amplitude of the vertical pattern. Then pixel math is used to add the amplitude to the dark bars. This allows deeper stretching and more detail in the faint fuzzy's.

Here is the webpage for downloading CCD Band-Aid. And of course, the program is freeware
http://bf-astro.com/ccdBandAid/ccdBandAid.htm

Regards,
Bob
http://bf-astro.com

[Juan Conejero]

Hi Bob,

Have you considered writing PixInsight module versions of your software? With our PCL development framework, you get direct portability to Windows, Mac OS X, Linux and FreeBSD without changing a single line of source code. PCL is an open source library released under PCL License 1.0.

[bob_franke]

Hi Juan,

That sounds good, except that I am not a "C" coder.

Cheers,
Bob

[Josh Lake]

Bob, as an SBIG STL-11000 user, thank you from the bottom of my heart.

It would indeed be wonderful to port this into PI, perhaps I'll dust off my old 'C How to Program' book...

[sreilly]

Bob,

Thanks for another PI tool. I haven't used this before with my camera and look forward to seeing the affect it has. Always looking for a better way.

-Steve

[Carlos Milovic]

Juan, we may use the code of the NotchFilter process (that is in your "review to-do list") to implement this. Is the same thing. Just have to automatically create the "mask". Not too much work.

[Carlos Milovic]

Now that I think about it... a better solution might be the new TGVRestoration process... It will allow to use maps in the frequency domain, so we should be able to correct these imperfections easily.

[bob_franke]

Hello Carlos,

Almost exactly 2 years ago you repaired the bars in one of my images with a Fourier transform. The result was impressive, but required experimentation. Search for the thread "Can PI fix my STL-11K vertical bars?"

I would like to propose a challenge to PI developers that will showcase the power of the software. This is to develop a filter that will determine the phase of the vertical bars and then lighten the dark bars to match the adjacent light bars. This process would be applied to all calibrated sub exposures before mean combining.

If done, this will be a great asset to many KAI-11002 users. This is not just a problem with SBIG cameras. I have seen this from other manufacturers. In Kodak's spec sheet, they suggest that the fix requires trial and error. This means testing and tweaking each camera.

Something like a notch filter is exactly what is needed to repair the vertical bars. The bars are 16 pixels wide and alternate light to dark across the entire frame. The amplitude seems to remain fairly constant across the image, but increases as the chip is deeper cooled. Binned images also have a higher bar amplitude. For a given camera, the phase of the bars only needs to be determined once.

Here is a link to the help file for my CCDBand-Aid program.
http://bf-astro.com/ccdBandAid/ccdBand-Aid.pdf

Here's a link to the Kodak paper on this topic.
http://www.kodak.com/ek/uploadedFiles/VerticalTimingOptimizatioInterline.pdf

Best regards,
Bob
http://bf-astro.com

[georg.viehoever]

Bob,
maybe you could post a couple of test files, so that people accepting your challenge can test their solutions.
Georg

[bob_franke]

Okay, test files coming tomorrow.

Cheers,
Bob

[georg.viehoever]

The explanation in the Kodak document, and also the image in your documentation seem to suggest that this is not a sine wave noise signal, but rather some square wave http://en.wikipedia.org/wiki/Square_wave. Did you ever measure the signal, for instance in an otherwise flat frame? Do you have an idea regarding its approximate form.

I am asking because an FFT based approach is good only if this is a sine wave.

Georg

[bob_franke]

Sorry Georg, it is definitely a squarewave. The bars are stronger with colder temperatures and longer exposures. The pattern is an odd beast. It does not show up in darks or flats. As I stated in my documentation, it is easiest to see in images of sparsely populated star fields.

Here are some images for development purposes.

http://bf-astro.com/3_Lum_subEx.zip
This zip file has three calibrated sub exposures. As scaled from the camera, the amplitude of the bars is about 1.20 ADU

http://bf-astro.com/32_mean_combine.zip
This file is a mean combine of 32 unregistered sub exposures. Click STF AutoStretch and you will clearly see the vertical bars.

Regards,
Bob

[georg.viehoever]

...I would like to propose a challenge to PI developers that will showcase the power of the software. This is to develop a filter that will determine the phase of the vertical bars and then lighten the dark bars to match the adjacent light bars. This process would be applied to all calibrated sub exposures before mean combining....

Bob,
you suggest that this would be applied to calibrated images. I wonder if these rather need to be raw (uncalibrated), or calibrated excluding flat division. Flats are there to adjust for optical effects, and not for sensor effects. I therefore think that the flat correction might interfere with the bar repair. What do you think?

Georg

[bob_franke]

Hello Georg,

That's a good question that makes a lot of sense. I have always worked with fully calibrated files. SBIG does not address this question in their CCDOPS help file.

I was just experimenting with a problematic set of images. Maybe I will get better results by only applying the darks.

Cheers,
Bob

[bob_franke]

Georg suggested that any process for correcting the vertical bars should be applied before the flats. I never thought about it, but he is correct. Just like darks, the band correction is an addition or subtraction process and should be applied before flats. So the question is, is it worth the inconvenience to do it correctly? Often times, with pretty pictures, it does not make a difference. Of course, with science, math processes must be applied in the correct order.

After some testing, I got inconclusive results. In my files, I found data for four sets of images that included flats, darks and the raw sub exposures. These images were also taken with a chip temperature of -25°.

I ran CCD Band-Aid twice for each set of sub exposures, about 30 for each image. This was after applying darks, but before flats and after applying darks and flats. In each case, CCD Band-Aid calculated a more aggressive correction when used before the flats.

Since the correction is added to the dark bands, the signal in all images increased. When checking the background areas, two of the images had an increased signal to noise ratio, and for the other two, it was reduced. The difference was small, only about 3 to 5%. In the case where the S/N decreases, CCDBand-Aid's result may have been an overcorrection. These results were determined with mean combines of dithered sub exposures.

Most importantly, the difference in the images was not detectable visually. In all cases, CCDBand-Aid effectively removed the bars.

Regards,
Bob