Shrinking source images size to effective size

[DmytroKushnirSoftserveinc]

Raw images after filtering have to mutch pixels without no information. We could crop border boxes containing an object of interest and therefore reduce the computational time wasted onto empty pixels crunching.

[DmytroKushnirSoftserveinc]

Proposed algorithm:
Having filtered image with pixels other than an object of interest set to 0 (NaN)

1. Compute border boxes of object at all dataset images (TODO : Or it could be improved further as we can optimize on level of single fragment?)
2. Select covering border-box, cropping out dataset of new image size.