Traffic image segmentation based on maximum posteriori mutual information

Abstract

Thresholding is the basic way for traffic image processing. Two-dimensional (2-D) thresholding methods can get better results. They used a threshold vector to divide a 2-D histogram into object, background, edge/noise three parts. Object and background parts were the common parts of grayscale information and spatial information. Mutual information focuses on studying the cross-section between entropies of two distributions, so it was considered to improve some disadvantages of the current 2-D thresholding methods. Experimental results showed that the proposed method could get better results.