Study of Local Binary Patterns

Abstract: This Masters thesis studies the concept of local binary patterns, which describe the neighbourhood of a pixel in a digital image by binary derivatives. The operator is often used in texture analysis and has been successfully used in facial recognition. This thesis suggests two methods based on some basic ideas of Björn Kruse and studies of literature on the subject. The first suggested method presented is an algorithm which reproduces images from their local binary patterns by a kind of integration of the binary derivatives. This method is a way to prove the preservation of information. The second suggested method is a technique of interpolating missing pixels in a single CCD camera based on local binary patterns and machine learning. The algorithm has shown some very promising results even though in its current form it does not keep up with the best algorithms of today.