Generating an Image from a Scene

[Cody Waldecker]

Hi all,

I think this forum might be a little dead, but does anyone know a way to take the pixel coordinates and magnitude from a randomly (or not randomly) generated scene and use them to produce an image of the stars that would match that orientation? Thanks.

[Andrew Tennenbaum]

It’s not quite dead yet haha, I do check in here from time to time. The software for image generation does not yet exist in the public version of openstartracker, but I have some matlab code that does this, which I’m working on updating & making a C version w/python bindings for publication. If you’re still interested I can send you a copy. 

Essentially, to generate this accurately, you need to know image noise, brightness of a star of known magnitude, and the width of a star in pixels.

With these three datapoints, you can generate images of any arbitrary star. The point spread function of stars can be approximated as a Gaussian, which makes it easy to analytically calculate the amount of light that hits each pixel. You then add either poisson noise or Gaussian noise depending on if you prioritize image generation speed or accuracy respectively

[Cody Waldecker]

Andrew,

Thank you very much for getting back to me. I would greatly appreciate your matlab code for generating the images. 

I was able to get something working and it actually solved using Astrometry (image attached), but it doesn't model the light captured well (ignoring noise for now). I took the shortest path and took the flux for each star as an estimate of brightness expected at each pixel before applying a Gaussian for Airy disk effects. I would love to see how you have it modeled because the sample images on the Github look much more realistic than mine currently do.

Thank you,

Cody