Enhancement: Eccentricity

[schackey]

What do you think about using eccentricity as opposed to minor length as a threshold in detection blocks?

I feel that then to rule out small stars one can use min_area, and using eccentricity rather than minor_length would then be a pure parameter to control the allowed ellipticity of stars. Using minor length seems to try to do both at the same time, and sometimes i then find it difficult to constrain the parameters correctly to only detect the stars i want. What do you think?

I will include it in the branch i'm building, if you like the idea!

[lgrcia]

Yes I think that's definitely a good idea. I thought this was well made but its not at all.

As you say: when using a PointSourceDetection block we should filter and get only truly point sources (based on eccentricity). This idea was implemented in the AutoSourceDetection block (see the algorithm here) but I don't see this enforced on any of the shape-specific detection blocks. With that we may be able to get rid of the minor_length.

Here is a proposed implementation:

• we replace minor_length by eccentricty_bound in the base _SourceDetection class
• we filter for sources in the clean method, so that only sources with an eccentricity within eccentricty_bound are kept

This way we inforce shape-specific detection blocks to be truly shape-specific :). What do you think?

[schackey]

Perfect, yes that's exactly how I have implemented it so far! I will clean it up and pull request soon.

[lgrcia]

That's awesome, thanks! 🙏🏼