Consider using a kernel density estimation instead of relying on standard peak detection on the luminosity data. Using the KDE will transform the image into a continuous pdf (component of whatever kernel posteriors are chosen, e.g. Gaussian) which makes peak detection much easier. Also introduces a pseudo smoothing parameter, h, which can be played with to find suitable results.
Consider using a kernel density estimation instead of relying on standard peak detection on the luminosity data. Using the KDE will transform the image into a continuous pdf (component of whatever kernel posteriors are chosen, e.g. Gaussian) which makes peak detection much easier. Also introduces a pseudo smoothing parameter,
h, which can be played with to find suitable results.