Research And Application Of The Algorithm, Based On The Proliferation Of The Border To Keep The Image

Image diffusion is widely used in underlying computer vision research as an edge-preserving smoothing method, which can smooth the image while maintaining or even enhancing the image structure, so as to enhance image features representation and ultimately helps to improve the high-level vision algorithms. The existing classical algorithms for image diffusion have their limitations. It seems difficulty to achieve a good balance between smoothing messy texture and maintaining edges, especially in low contrast image regions. Besides, manual tuning of kernel parameters is often needed to achieve good diffusion effect.In this paper, based on the analysis of the existing algorithms, we propose a new edge-preserving based adaptive 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 making sure image boundary not be destroyed. The intervening contour measures whether the two pixels are distributed on both sides of the boundary, so as to ensure the diffusion process not to cross the boundary. The color component ensures the photometric similarity between two pixels by measuring their distance in range domain. At the same time, we apply a pixel-segmentation-property based approach in our experiment for adaptive kernel parameters selection, which improves the accuracy and adaptability of smoothing, helps to preserve edges in low contrast image regions, and avoids the complicated manual tuning process. Experiments show that the adaptive ICC method not only has better performance of edge-preserving and messy texture smoothing than bilateral filtering and mean shift, but also has faster diffusion speed.The application of image diffusion method is also studied in this paper. Due to the good performance of messy texture smoothing, the adaptive ICC method can be used in edge detection and content-aware image resizing as a pre-processing method, which successfully weakens the side-effect of messy texture on image features representation, and significantly enhances the processing effects of texture images.