A Gradient-based Coherence Enhancing Diffusion for Road Surface Image Enhancement

In road surface images, in order to detect the cracks that are very tiny, we have to enhance them first. The image enhancement is a new type of diffusion process that simultaneously enhances, sharpens, and denoises images. Conventional coherence enhancing diffusion enhances the flow-like cracks but the other unwanted elements on the road surface are enhanced at the same time. Also it doesn´t sharpen the edges of cracks. In this article, we propose a gradient-based coherence enhancing diffusion that enhances the cracks and also eliminates the other unwanted elements. The new diffusion distinguishes cracks from unwanted elements by gradients on the assumption that the gradients of cracks are approximately unchanged, and takes different strategies to control cracks and other elements. The new approach also absorbs the idea of forward-backward diffusion to determine the strengths and directions of the diffusion process in order to sharpen the edges of cracks. As a result, both the edges and the flow-like structure of cracks are enhanced. Theoretic analyses and experimental results both support the effectiveness of the approach