Finally we will look at rank order filtering (rank) which has been suggested as a pre-processing step for stereo algorithms. Load in both stereo images of the lego house (one at a time) into the image calculator.
Apply the rank filter to both images, the range parameter should be left at 5 (which specifies an processing window). You will need to set the precision parameter to approximately 4 times the estimated level of image noise. Try to pick a sensible level that eliminates noise but retains discriminable features.
Can you say anything regarding the apparent similarity of grey levels for equivalent locations in the original images and the rank filtered images?
Is this a sufficient measure of performance for selecting a suitable pre-processor for a stereo algorithm?