PDA

View Full Version here: : PI - combining images from multiple nights


luka
12-07-2018, 10:14 PM
Hi,

I have been following the lightvortexastronomy (LVA) tutorial (http://www.lightvortexastronomy.com/tutorial-pre-processing-calibrating-and-stacking-images-in-pixinsight.html#Section9) for pre-processing of images. I have a question about how to deal with data from multiple nights.

For images from a single night the LVA tutorial suggests the following:
1. Image calibration
2. Cosmetic Correction
3. DeBayering
4. Optimisation of weightings of images using SubframeSelector script.

followed by
5. Star alignment
6. Local normalisation
7. Image integration
8. Drizzle integration

So far so good. Now, my question is at which step to combine data from multiple nights.

LVA suggests to do steps 1-4 for each night and then to combine all data starting with the star alignment (step 5). Now this approach will calculate individual weightings (between 50 and 100) for each image for each night before the combination. The best images for each night will be the highest ranked. However, if one night was particularly bad (windy, cloudy etc) the images from that night may be weighted as high or even higher than much better images from the other nights.

Wouldn't it make sense to combine images from all nights in the SubframeSelector and weight them all based on their quality compared to each other and not "separated" by nights?

Thanks in advance.

Atmos
13-07-2018, 05:09 AM
If I was to take images over the course of a week I would end up with a big folder of “Calibrated” images. I use Blink to go through and remove any bad ones and then treat that whole calibrated folder as if it was just one data set. Doesn’t matter if they were taken in one night of multiple, the same routine.

Slawomir
13-07-2018, 05:29 AM
Hi Luka,

I always do calibration (step 1 in your workflow) on the whole lot of subs from multiple nights and use subframe selecector on the whole lot to identify worst and best subs and then do a visual check with blink tool. Once you identify the best sub I use it as a reference for integration of all subs. (after doing a cosmetic correction and star alignment) I think this method is potentially more robust than doing separate lots for each night.

luka
13-07-2018, 05:35 PM
Thanks guys, that was exactly my thinking. And thanks for the tip to do steps 2 & 3 for multiple nights as well, I didn't think of that.

Suavi, you said that you do calibration on images from multiple nights. Do you have a fixed setup (observatory)?
I have to remove scope/camera from night-to-night (portable setup in the backyard with kids) and the flats will change from night-to-night. Hence step 1 has to be done separately for each night.

Octane
13-07-2018, 08:48 PM
In that case:

Calibrate each night's worth of data with their respective flats.

Then, register and combine as per usual.

H

Slawomir
14-07-2018, 05:19 AM
Luka - I also need to set up for each session, but I keep my camera attached to the telescope so the same flats can be used for subs from multiple nights.

sil
17-07-2018, 08:51 AM
I generally use what you originally listed (steps 1-4 for each set, the step 5 to create a set of registered images from ALL sessions combined, integrate etc..). I should but dont use blink as my sets are around 500x 56MP RAWS each session and i'm generally lazy and happy enough with my results regardless. Except if its rare data for me to capture then I put in the time to manually check each frame to try to make the most with what I got.

I capture untracked dslr and i'll often find five good frames that have my target well centered for least distortion and manually align them (dynamic star alignment) then integrate them to an integration target frame to then register the whole set against. When I have noisy (high iso) data I find this really helps registration accuracy for me. Plus I can easily combine data from different equipment if I want (ALL frames are scaled and cropped to the same field of view and size of my integration target).