Improve correct_image curve fitting by rejecting altitude bins with few images

[cappelletto]

After computing altitude histograms (e.g. default 25 bins), some of the bins end with very few images. The curve fitting process is affected by the uneven distribution of image samples per altitude bin. This tends to produce visible artifacts when developing (processing) the images.

A potential solution can be defining a threshold for minimum number of images per bin to accept that bin in the curve fitting step.

[cappelletto]

A sample scenario:

INFO ▸   Bin 07 ( 4.2 m < x < 4.3 m): 0 images
INFO ▸   Bin 08 ( 4.3 m < x < 4.4 m): 5 images
INFO ▸   Bin 09 ( 4.4 m < x < 4.5 m): 4 images
INFO ▸   Bin 10 ( 4.5 m < x < 4.6 m): 30 images
INFO ▸   Bin 11 ( 4.6 m < x < 4.7 m): 79 images
INFO ▸   Bin 12 ( 4.7 m < x < 4.8 m): 255 images
INFO ▸   Bin 13 ( 4.8 m < x < 4.9 m): 527 images

Bin 09 could be rejected due to the relatively low number of images (4)