Reported coordinates seem shifted

[koszullab]

Reported coordinates seem shifted

[cmdoret]

Hi !
Glad to know it's useful :)

Yes sometimes the pixel with the highest correlation with the kernel is next to the pixel with the pixel with the highest contacts. Reducing the kernel size using a smaller --win-size value can help (correlations are computed using a smaller neighbourhood, so distant pixels are less likely to interfere). This might also have the side effect of increasing the number of false positives.

Coordinates very close to the diagonal can also be shifted. This happens when the kernel overlaps the lower triangle. This is fixed using the --full option, which applies a "missing mask" taken into account when computing correlations. Pixels below the diagonal, or in missing bins are excluded from the calculations. This option slightly increases run time and memory usage, but it seems to yield more accurate results, so it might become the default in version 1.0.0.

Thanks for the feedback !

[Phlya]

Thanks, I'll explore these options!