The image resolution is reduced by combining square "bins" of pixels into one pixel. For a reduction of 2X, every pixel in the reduced image is computed from a 2x2 bin of pixels in the original image, so the reduced image has one 1/4 as many pixels. Reduction levels of 3X and 4X use 3x3 and 4x4 bins and reduce the image size even more.
Resolution reduction can be used to reduce vision processing time at the expense of precision. Many vision applications can be done with acceptable accuracy with a reduction of 2X or more.