PSF Resizing, Global Backgrounds

U-Band PSFs

So the ePSFs that we have been getting for u-band, as you know, are being oversampled and thus are not the right size (and the pixels are not the same size as the input images). After spending some time fruitlessly trying to resize the PSF directly into the builder (and to no success), I turned to rebinning. The methods that scipy use did not remotely work, but then I realized that the PSFs are images and you can use image resizing procedures. As it turns out, scikit-image has an image resize function that uses interpolation to change a 2D array to any size you want. Below is a simple example:

This SHOULD be fine so long as resizing the ePSF back to the original image shape returns the pixels to the right size. There is no reason for me to think this would be untrue.

Global Backgrounds and Subtraction

As Marcin predicted, there are no clear 2 peaks for the asymptotic values. I tested three different methods of measuring an average: median with no alterations, sigma-clipped median, and weighted mean. I sigma clip with a bias towards larger SB values, as we know the lower values are definitely not ideal measurements. These results (SB histograms and medians) are below:

All pretty consistent values! I use the sigma-clipped value to generate a new global background subtracted profile in the matrix plot figure (the orange curve). The background measurement now has an associated error, which is determined as the standard deviation of SB values where the smoothed gradient is close to 0. I also have, after far too much time, added a legend to the plot to explain what the curves are. Finally, I fixed the label on the top row to have the same font size.

They definitely don’t match up identically to the green profiles in every bin, but they look far better than I expected for such a rough measurement (of both the background in the green profiles and the new orange profiles). We definitely lose some data to negative values in some bins, but that is to be expected at such low SB values.