For some data, the initial Otsu thresholding at 1.5 doesn't work properly. What if we do a precorrect0 at 0.5 first?