incorrect disparity range manipulation

[gfacciol]

This line (and surrounding code) is in charge of extending the disparity range for the stereo matching:

s2p/s2plib/rectification.py

Line 194 in 942de31

 disp_min *= (1 - np.sign(disp_min) * cfg['disp_range_extra_margin']) 

But these manipulations are incorrect. If disp_min=0 to begin with, then no multiplication will ever increase the disparity range. The correct way to increase the disparity range is with:

dmin-= (dmax-dmin)* cfg['disp_range_extra_margin'])
dmax+= (dmax-dmin)* cfg['disp_range_extra_margin'])