weird mean after stacking

[rxdeath]

whenever i read other peoples questions on here, i feel like a complete moron since mine seem so simple by comparison but here goes:

i have 5 red subs i took all with mean of around 590 in 16bit space. 

when i stack them, the mean drops to 280

for the 5 blue and green subs done at the same mean of 590, they stack to around 590.  not sure why red is different, i understand exposure wise it's different which is why i use the background mean to ensure i'm getting a even skyfog level across all subs.  asi1600mm-c used for all exposures, at same gain/offset, etc

what's happening?  also i know that isn't enough exposures to get a solid picture, i'm just experimenting find the right exposure for the sqm in my backyard and learning how to evaluate proper length subs

i can post them, but its like 150mb of subs, and i can ensure using statistics window and going through them the mean is similar and then drops in the integrated frame, while green and blue do not drop

thanks pixinsight ninjas

[RickS]

It's probably a consequence of the normalization done by ImageIntegration. You may get different results if you choose a different reference image (the first file in the list of input images) but the important thing to understand is that it doesn't actually matter.  When you combine the colour masters and do colour calibration the values in the red master will be scaled to give a correctly balanced result.

If you still find this disturbing you could always use one of the red subs as a reference image and apply LinearFit to the red master.  That will give you a mean value more like you're expecting but it won't make any difference to normal processing.

Cheers,
Rick.

[rxdeath]

thank you once again.  our previous deal gave me the skills to find this stuff out, but apparently i haven't come far enough along on image processing to know this doesn't matter.  just for s&g i will try a different reference image.  i haven't used linearfit or colorcalibration.   right now i'm following the lightvortex tutorials, but my first m51 image look like yellow crap.  the subs aren't incredibly terrible considering my lp zone.    i have been pounding through warrens book, i'll keep my nose to the grindstone, thanks a lot rick

[RickS]

No problem. It takes a while to figure it all out Fighting with LP is always tough unless you go narrowband.

[ceteris_paribus]

I have this same problem as well, and I was wondering if anyone actually knew why this happens.  Let me outline quickly my problem as it might differ slightly from the original poster's.

I have about 30 registered frames, all but one of which have some sort of black border due to registration.  About 1/3 of them, when used as a reference frame bring the median up by about a factor of 4.5x (not all 4.5, some 4.2, some 4.3, etc.).  Some, however, leave the median the same.  To debug, I was using only Additive normalization, no scaling, no rejection of any sort, so aside from averaging the frames, there should not be any further calculations except the normalization. 

The additive normalization formula looks to take the pixel value from the current frame, subtract the current frame's median, then add the reference frame's median.  I did the calculations manually for a number of pixels, and there was no way that I could reproduce even close to a 4.5x increase in value of each pixel.

I suspect that this has something to do with the registration's 0 value pixels around some of the edges, but I cannot see how and why.  I cropped the same images and they stacked exactly as would be formulaically expected.  Even when I clipped the 0 value pixels out, the median of each image did not change very much; certainly not enough to increase pixel value by 4.5x.

I am aware that, at this point, I can just use a different reference frame to make everything more to my liking, but I am now curious as to what is going on under the hood.  If anyone is curious, I have included a link to a number of the frames in my Dropbox.

https://www.dropbox.com/sh/6eeh8dwzt3x5904/AAA16HxeF0aBSl7Y8ISONPpXa?dl=0

Thanks in advance,

-John