4x4 Census Transform
Olivier Rukundo · 2020
This paper proposes a novel 4x4 census transform (4x4CT) to encourage further research in computer vision and visual computing. Unlike the traditional census transform (3x3CT) which only uses a nine pixels kernel or 3x3 window size, the proposed 4x4CT uses a sixteen pixels kernel or 4x4 window size with four overlapped groups of 3x3 kernel size. In each overlapping group, a reference input pixel profits from its nearest eight pixels to produce an eight bits binary string convertible to a grayscale decimal integer corresponding to the 4x4CT's output pixel. Preliminary subjective assessments were more promising with the 4x4CT transformed images than 3x3CT transformed images.