Fast Block Edge Direction Analysis in DCT domain

Abstract

Detecting edge existence and analyzing the edge direction in a block level is useful in the applications, such as perceptually improved image processing, blocking artifacts reduction, content-based retrieval, senses changing detection, etc. This paper proposes a novel Fast Block Edge Direction Analysis method. The method is derived from a simple pixel domain in a deductive way. Then we demonstrate how to calculate the key parameters of the method with DCT coefficients. With this method, one block is classified into one of the 17 category edges: No edge plus 16 different edge patterns. The calculation of the key parameters can be calculated with extremely low complexity both in pixel domain and in DCT domain.