Anisotropic diffusion filtering method with weighted directional structure tensor. (August 2019)

• Record Type: Journal Article
• Authors: Wang, Yu; Wang, Yuanjun
• Is Part Of: Biomedical signal processing and control. Volume 53(2019)
• Publication Date: 2019-08
• Subjects: Medical image -- Weighted directional structural tensor -- Directed diffusion tensor -- Anisotropic diffusion filtering; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2019.101590
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: An improved anisotropic diffusion filtering method is presented. The directional structure tensor (DT) replaces the traditional structure tensor (ST). The weighted directional structural tensor (WDT) is constructed by combination of DT and non-local mean. The new diffusion coefficient constitutes a directed diffusion tensor. This method can process a variety of medical images. Abstract: The anisotropic diffusion filtering algorithm has excellent smoothing performance for medical images, but the normal diffusion filtering algorithm will blur the edges and details. In this paper, we construct weighted directional structural tensor (WDT) and propose an anisotropic diffusion filtering method based on the WDT to overcome the fuzzy appearance drawback. The proposed algorithm first constructs the directional structure tensor based on the traditional structure tensor, and then add the definition of the weight in the non-local mean to construct WDT. To further protect the edges and small structural features of the image, we also set the diffusion weighting coefficient according to the eigenvalues of DWT to construct directed diffusion tensor. Experimental results indicate that the proposed method shows better performance than other filtering methods, and greatly improves the image sharpness, presents better image details and maintains the edge contours while denoising.