PixInsight Forum

PixInsight => General => Topic started by: alanrock on 2019 October 14 18:19:45

Title: RANSAC Issues--they're back!
Post by: alanrock on 2019 October 14 18:19:45
Trying to stack some subs using BPP and it fails with RANSAC errors, so I tried to just register the images with the Image registration process, and I get:
Error: Unable to find an initial set up Putative star pair matches.

I have tried Increasing RANSAC Tolerance as high as 7. I have tried increasing and decreasing Log(sensitivity). I have tried Noise Reduction as high as 3.
Ran the files though DSS and they registered and stacked just fine.
Any ideas? It's not a super contrasty image, but there are plenty of stars, which DSS apparently was able to find.
I'm using 1.8.7 on Mac OS X 10.15

Title: Re: RANSAC Issues--they're back!
Post by: airscottdenning on 2019 October 14 20:46:03
I had a similar problem, and Juan diagnosed it as having to do with the new PSF-based alignment algorithm  From his response:

new version of StarAlignment performs PSF fits by default for all detected stars. Because of this feature, the new version is much more demanding of good star shapes. Accurate PSF fitting is necessary for the new distortion correction algorithm, which aims at achieving centipixel registration accuracy, even under strong nonlinear distortions.

The problem with these images is that they are out of focus. As a result, PSF fits cannot be as reliable as the tool requires, and hence most stars are being rejected. To solve this problem you have to disable PSF fits, which returns the tool to a behavior similar to previous versions:

- Open the Star Detection section.
- Set Compute PSF fits to Never.
Title: Re: RANSAC Issues--they're back!
Post by: alanrock on 2019 October 14 21:07:27
Well, PSF wasn't the problem, but it certainly got me thinking. I thought my focus was decent, so I went back and looked at the frame I was using as a reference, and it turns out that it was a frame I had meant to discard--let's just say it was a mess. When I deleted it and used the frame I had actually intended to use, everything worked just fine!
Title: Re: RANSAC Issues--they're back!
Post by: ngc1535 on 2019 October 15 00:22:27
Hi Alan,

But... (just in case others read the final conclusion to your issue) the PSFs of your reference *were* the problem!
Sure it was an accident you picked a poor quality/non-image and the PSFs in the other images are OK...but the algorithm did what it was supposed to and could not match PSFs. You also made the statement that you had no problems in a different program...but is it likely any other program would have encountered the same issue?

Anyway, the fact that it was not matching was actually helpful for you to figure it out. Perhaps setting match PSFs to never might make things more confusing for an image that is a "mess."

Is this a fair way to tie up the thread?