Edge-preserving based adaptive ICC method for image diffusion

Image diffusion is the underlying machine vision research method, and is frequently used to remove image noise. In this paper, we propose a new kernel function for image diffusion by combining the intervening contour (IC) and the color (C) component, called ICC kernel function, which can smooth inner messy texture while maintaining image structure. The intervening contour is used to ensure the diffusion process not to cross the boundary, and the color component ensures the photometric similarity between two pixels. A pixel-segmentation-property based approach is applied to adaptively select the kernel parameters. Experiments show that the adaptive ICC method has better performance of edge-preserving and messy texture smoothing than bilateral filtering and mean shift. Our adaptive ICC method enhances image features representation and helps to improve high-level recognition algorithms.