Novel Approach to Pavement Image Segmentation Based on Neighboring Difference Histogram Method

Conventional human visual pavement distress detection method is very costly, time-consuming, labor-intensive, and is often dangerous due to exposure to traffic. It possesses various drawbacks such as being unable to provide meaningful quantitative information and with a long periodic measurement. In this paper, a novel pavement image-thresholding algorithm based on neighboring difference histogram method (NDHM) is proposed. The main idea of the proposed method is based on the facts that: (1) the distressed pixels in pavement images are darker than their surroundings and continuous; (2) the thresholding value is strongly related with the image standard deviation. In this method an objective function for maximizing the divergence between the two classes is constructed. The paper compares the new method with the classical discriminant analysis method of Otsu and the entropic method of Kapur et al. The experimental results have demonstrated that the distresses are segmented from the background correctly and effectively.