I am trying to use the Lenstronomy lens fitting library to fit a simple SIE model to some lens images. The images are from SDSS and Legacy Surveys data releases so they have a lot of background noise. I have some masks that cover the arcs for each image so I am able to isolate the arc and the center and supress all the background, but what happens is that when I look at the K-Convergence Map that gets outputted, the center is near a corner and the Einstein Radius is much larger than it should be.
I have never really used Lenstronomy before and I am relying heavily on the tutorial docs in the github page. I think the problem is something to do with the way I pass in an image and the noise mask.
My main questions are just:
- Should I pass in the entire grayscale SDSS image or mask out the lens center and arc?
- Should the noise mask be highest in the background or in the section with the lens center and arc, or is it something else?
- Is it possible to restrict the center? I have already tried this but if I do it then the PSO fitting just “gives up” and returns the initial guess parameters.
- How sensitive is PSO to the initial guess? Does it need to be super accurate or can I just expect a good PSO to fit to the proper parameters?
Thank you for all your help.