Anisotropic Diffusion Image Denoising Method Based On Bilateral Structure Tensor

With the development of digital processing technology,digital image processing technology is widely used in the fields of pattern recognition,Resource Planning,Urban Planning,and biomedicine.However,in the daily research practice,due to the limitations of the outside world and its own conditions,the image will be affected by various types of noise during the process of acquisition,transmission and storage,resulting in the reduction of image quality.This result has caused great problems in many high-precision research fields,resulting in limited subsequent processing.Therefore,how to effectively denoising the noisy image so as to obtain more accurate and reliable information from the image has become a research hotspot in the field of digital image processing.In the process of image denoising,because the edge and texture information belong to the same high-frequency information as the noise,how to balance the noise filtering and the edge texture information keeping has become the focus of image denoising research.In recent years,the diffusion method based on partial differential equations has become a new tool of image filtering due to its anisotropic diffusion and other features.This method can selectively smooth the image while filtering the image noise,so as to perform diffusion filtering in the smooth area and detail protection in the edge area,the contradiction between filter denoising and edge protection is well resolved.Based on the Weickert diffusion model,this paper introduces the image diffusion filtering method based on structure tensor.On this basis,the traditional structure tensor diffusion filter needs to undergo two times Gaussian kernel convolution in the process of image denoising,which does not make full use of the image structure information,so introduces the bilateral structure tensor into the diffusion model to construct a new diffusion tensor for image denoising.Different from the traditional structure tensor,the bilateral structure tensor is based on the idea of bilateral filtering algorithm,which is constructed by considering the similarity of the spatial position information and local gradient intensity of the image pixel,thus it has the ability to describe the structure details more accurately,and saves the twice Gauss convolution process during the construction of the traditional structure tensor,and protects the image details better.Using MATLAB to carry out simulation experiments on the proposed image denoising algorithm and making a contrast analysis through the objective performance parameter analysis and the subjective vision,the result indicated that,compared with the traditional structural tensor method,the bilateral filtering and the non-local means filtering denoisingalgorithm proposed in recent years,the algorithm proposed in this paper can effectively remove the noise while maintaining the texture feature information of the image.