Hi Wade,
The reason is here:
** Warning: No significant data in MRS noise estimation routine - using k-sigma noise estimate.
For some odd reason the multiresolution (MRS) noise evaluation routine has been unable to compute a valid noise estimate for the final integrated image. When this happens, a second algorithm is used (k-sigma) which is less accurate than MRS but cannot fail.
Obviously the k-sigma estimate is too high: 1.902e-3, which is nearly the same value computed by MRS for one of the raw images. For this reason the final SNR improvements are too low, and are not reflecting the true improvement achieved.
MRS noise evaluation should not fail under normal conditions. Do your images have strong gradients? Gradients can be detected as significant image structures sometimes, and when this happens the MRS routine may be unable to find any noise throughout the whole image. I have some ideas to fix this problem; perhaps I'll implement them in the next version of ImageIntegration.
Another possible cause is cold pixels. You could also try to calibrate the images with PixInsight, to see if our automatic dark scaling routine goves you better results.
A workaround: select the integrated image, and run the noise evaluation script. If it gives you similar problems, try defining a partial preview and running the script on it. When you get a good MRS estimate for the integrated image, simply divide the estimate for your reference image (1.967e-003) by it, and you'll get the SNR improvement.
And ... I'd like to see the images because they may help me to improve the noise estimation routines
