Why align the mean ±3× standard deviation of RGB scores with the mean ±3× standard deviation of SDF scores?

[limaodaxia]

To calibrate the distributions of scores, we align the mean ±3× standard deviation of RGB scores with the mean ±3× standard deviation of SDF scores by applying an affine transformation to the RGB scores.

Hello @jayliu0313 , where you got this idea?

[jayliu0313]

Due to the significant numerical gap between these the image feature and the SDF feature, we use statistical methods to align them, ensuring that scoring is appropriate between them.

[limaodaxia]

for L in range(sdf_c.shape[0]):
    tmp_map[index[L]] = sdf_c[L]
    if(s_map[index[L]] == 0) or (s_map[index[L]] > sdf_c[L]):
        s_map[index[L]] = sdf_c[L]

when get the score map of sdf, why use the min val of distance to update the map?