Using DBE to remove background?

[georg.viehoever]

Hello,

I just shot a couple of pictures of M31, and stacked them in DeepSkyStacker before moving to PixInsight. Since my FOV is smaller than Andromeda, I used DeepSkyStacker's mosaic mode. The histogram stretched result is attached below. Obviously, I have a problem with calibration. I thought it might be possible to get rid of the background gradiants using the DBE process, but no matter what I do, I still see the rather prominent lines where the pictures begin to overlap. Any idea how to handle that (short of taking the pictures again  :wink: )? I know this is more a matter of desaster recovery than of image enhancement, but anyway...

http://cid-c56e60845cfa6790.skydrive.live.com/self.aspx/Pixinsight/andromeda%20combined_stacked2.jpg (2.7 MBytes)

Kind regards,
Georg

[twade]

Georg,

You need to flatten each image individually before you mosaic them together.  If you flatten them correctly, they should be seamless.  

Flattening an image properly is an iterative process.  I generally find areas where I know there are no nebulae, dark or dust clouds.   In addition, I try to keep the number of sample points to less than 20 and spread throughout the image.  After the first round of using DBE, I create a very aggressive screen stretch.  This will show me any remaining gradients.  I will then add a sample points to these areas.  You should save each DBE processing icon.  Sometimes, I find removing some while adding others works the best.  Your resulting FLAT should mimic your gradient.  You should not have hot spots are dark spots in the FLAT.  If you do, you need to remove the particular sample point that is causing the bogus result.  I will then go through the process again until all the gradients are gone.  I then repeat this process for each frame.  It's time consuming but well worth the effort.

Wade

[georg.viehoever]

Wade,

thanks for the valuable hints. I was indeed able to create a much better version now.

http://cid-c56e60845cfa6790.skydrive.live.com/self.aspx/Pixinsight/andromeda_mosaik_4.jpg (1.9 MBytes)

It is still a far cry from other results shown in this group, but for me it was still surprising to see what could be done with with such raw material. Here, for the benefit of others, the essential steps that I did.

1. Stack both pictures separately. Apparently, darks/flats are just not enough to level out the differences of pictures taken separately (and on top of that, the flats I took during this night appear to have a problem...). This was easily done in DeepSkyStacker.

2.  Process both pictures separately, except for the very last step

3. Find and read the DBE documentation. You'll find it bits in several posts in this forum, but the most complete set appears to be in a tutorial: http://pixinsight.com/tutorials/LE/DBE-example/en.html

4. Do the DBE. I used 20 samples per row, reduced the smoothing factor to 0.05 and set the sample radius to 11. With more smoothing or less samples, the background just would not go away.

5. As suggested by Wade: Stretch the result and inspect it for gradients. In my case, there were several "holes" in the result, caused by DBE samples in the vicinity of bright stars. Moving/deleting those points helped. This is indeed an iterative process until you find the best set of samples.

6. Histogram stretch, HDR Wavelet, ACDNR

7. Final Histogram stretch, taking care that  both images have a similar appearance in Background, Color, Brightness.

8. Combine images using the Beta StarAllignment process in Mosaic mode, then some final retouching in GIMP.

I guess I could produce an even more pleasing result if I processed luminance separately, which I will do when time allows. But there is astrophotography, and then there is the rest of the life that somehow also consumes time...

Thanks for your help,

Georg