Deceived bilateral filter for improving the classification of football players from TV broadcast
Saul Calderon Ramirez, Francisco Siles-Canales · 2014
This paper presents a novel image abstraction technique, the deceived bilateral filter (DBF). The DBF combines border sharpening with simplification of homogeneous regions to achieve moderate de-blurring and colour emphasis. The aim of the DBF is to remove noise of colour and shape descriptors (histograms and contours) of the blobs corresponding to football players. The experiment implemented for this paper consists in an unsupervised histogram based classification of the blobs based on the team of the football players aiming to evaluate histogram de-noising. The results show an average classification success rate improvement of 11%.