Hello, I'd like to face-detect and subpixel-interpolate at the same time to check specifically the depth of face centers. When I enable depth alignment however, things become very laggy. Is this performance degradation due to computational complexities of the alignment or due to the 16bpp integer disparity buffer being unfriendly to scale and adjust efficiently? If so, is there a way to alternatively get a transformation matrix that lets me map single pixels from color/face to depth instead?
