Reflections Removal Using K-Means Clustering

Abstract

The reflection generated from the semi reflectors (glass windows) affects the algorithm efficiency of the computer vision. It is a naturally hard problem to separate a single overlapped image into a reflection image and a transmission image, which is a kind of blind separation problems. This paper presents an automatic algorithm to separate the reflected and the transmitted components of a single superimposed image by clustering mixture pixels. Various examples are tested to validate the effectiveness of the proposal algorithm. Experimental results proved that the proposed algorithm can produce acceptable results.