Wednesday, 13 November 2013

Massively Improved Flatfield System

The telescope has had flat fielding built into it for some time now in order to improve the quality of images we return. Today I fixed some problems with how those flat-fields are applied, which will improve images enormously.

I have never been entirely satisfied with the flat fields on Galaxy Cam. We have used flat fields created by us (by Scott many years ago and Dan since) as well as using flat fields submitted by users (as is currently the case on cluster) but there has always still been some vignetting on the images.

Yesterday, I decided to have a go at making my own flat field. I decided to use the raw night-sky images taken by the telescope to generate this flat field, as we have no shortage of those! 

I initially used 20 images of different radec coordinates from the past two weeks, combining them in a median stack to get rid of the stars so that only the background light remained. When I tested this flat field on the site, the result was not as good as I had hoped for. The field was still not flat and it was barely an improvement on the flat-field we had before. I increased the number of images to 40. Still the result was not any better than previously.

This morning, I decided to up my game a bit. I used 160 images of different targets from the past month, again median stacking them. The flat-field itself looked perfect and noiseless, but when applied to images it still wasn't looking great.

I decided to do some investigating and to cut a long story short I discovered to my surprise that the algorithm being used to apply flat-fields to images was not doing so correctly - in fact it didn't even look like a flat-fielding algorithm at all. While it did improve images over no flat-field at all (somehow), it was still way off track.

I updated that algorithm and tested it out with my new flat field and the results are absolutely remarkable. What is more, because the flat-field is applied by the site every single time you look at an image (not just when it is taken), images far back into the past also see the same improvement. When I went back over images that I had previously discarded as being of objects too faint to see in a 3 minute exposure I suddenly found them beautifully clear. It is almost like having a whole new telescope.

The following are recent images are of the whirlpool galaxy, which is barely observable at this time of year, and therefore poses a real challenge for flat fielding. A better placed target really comes out of the other end looking fantastic. So now would be a good time to go over some of your images again. You might just find them looking a lot better than you remember!

'unprocessed' image isn't promising

Flat-fielded with old algorithm

With the updated algorithm