help with pixinsight with plenty lrgb and no darks or bias frames

[iwols]

hi read/watched hours of pi demos but still not sure where to go, im trying to process 6x5 min lum frames,7x 5 mins red,6x 5 mins green and 6 x 5mins blue frames of the bodes galaxy without bias ,flats or darks in pix insight manually, and hope someone can help me along the way,so far ive opened all the images,opened star alignment put all the files in there and pressed apply global(with the default settings) and saved all the files to a folder called reg star aligned.then opened all the lum files from reg star aligned folder into image integration process window and changed pixel rejection setting to winsorized sigma clipping and ran, saved the new integration 1 image(not high or low rejection image)to a folder named integration and named bodes L,did the same with the r,g,b channels and saved the image integration images respectively as 16 bits,opened channel combination  and combined the rgb image into one,stretched and finished up with the image 06 which is funny coloured what part have i missed or done wrongly,ive uploaded the original lights folder,the reg star folder the integrated folder and the final image 06 if anyone can help thanks guys heres  the link 

https://www.dropbox.com/sh/vdnmonjwwx846ux/AAD_QO0jyvSho-5rndttrRRVa?dl=0

and a jpg of the final image

[brrttpaul]

I am not sure new to PI myself but that looks about right if you havnt taken any darks or bias, have you done auto dynamic background extraction? it should take away some of that gradient. I think you need to keep it linear though before you use it but as I said im a beginner also. I do STF then ABE then colour neutralisation, colour calibration, then switch off STF and do a histo stretch, curves, then noise reduction

[Star101]

I only use OSC camera to image so don't know about LRGB.

To me, that image looks like what I would expect when no Bias/Darks/Flats have been applied. The colour gradient looks like light pollution. If you created some Bias ( and Darks if need be). You could use Bias as Darks by renaming the Bias file to something such as Bias as Darks.( Credit to ollypenrice for that idea ). If you want to save time. Then run again with those calibrated in. You would probably get a much better image.

[wimvb]

You forgot linear fit. Choose one of the r, g, or b images (lowest noise, eg) as reference and apply LinearFit to the other two. Then combine, dbe, background neutralisation and colour calibration.

Good luck

Ps

Don't forget to crop before dbe

Btw, with fewer than 10 frames per image/channel, you should check if min/max or percentile clipping gives better results than sigma clipping.

[iwols]

thanks all does anyone fancy a quick go with my data to prove how bad i am at the minute ,going a little bit crazy and not sure what the goal is ,or if my data is any good ,hopefully iwols

[wimvb]

Linear fit with bodes_g as reference.

RGB channelcombination

DBE with tolerance = 1, sample size = 15, subtraction with normalisation on. Applied twice, this gets rid of most of the gradient.

Normally I wouldn't use that many samples, but you had various colour gradients, and there is no faint dust to keep (chasing IFN in this image will not be an option).

(It's possible to remove most of the gradient with less samples. Place a sample in each corner, one or two more along each edge, and about 6 or so spread evenly over the rest of the image, avoiding the galaxy and stars. Apply DBE twice.)

Background neutralisation with Preview01 as reference

Colour Calibration with Preview01 as background reference, and Preview02 as white reference, no structure detection

These processes reveal Holmberg IX as a very weak patch in the background. Your image is quite good but noisy because of the low number of frames.

Your smaller stars are undersampled, so you could try using drizzle integration once you've gathered more data.

And since many stars are undersampled, doing deconvolution probably won't improve much.

iwols_pixinsight.xpsm

Ways to improve:

Try to avoid problems before they arise:

- collect more data

- use calibration frames. Shoot a number of dark frames at the settings you used for this image, and examine for hot pixels and amp glow. Since the gradient is along the top and bottom edge, amp glow can't be ruled out.

[Oddsocks]

On 17/04/2017 at 01:33, iwols said:

read/watched hours of pi demos but still not sure where to go

 

On 20/04/2017 at 15:09, iwols said:

does anyone fancy a quick go with my data to prove how bad i am at the minute ,going a little bit crazy and not sure what the goal is ,or if my data is any good ,hopefully iwols

As Wim posted, you need many more frames to reduce noise and you must calibrate your images. If you don't start with good data then you are trying to post process with both hands tied behind your back!

This was the best I could do after dropping some of the frames that showed trailing.

After alignment and combination the process sequence followed was exactly that shown in the two PI demonstration videos for NGC 1808 with two additional steps taken at the end, a range mask to add selective ACDNR noise reduction, a little for the galaxy then invert the mask and add a lot of ACDNR to quieten the background, followed by an inverted star mask and colour saturation boost for the red channel only to bring out some of the red stars but leave the galaxy and background alone.

You can find the two PI videos that were followed here:

http://pixinsight.com/videos/NGC1808LRGB-vperis/index.html