Research On Low-Dosed X-Ray Computed Tomography Imaging By The Fractional-Order Perona-Malik Diffusion

After more than 30 years of rapid development, Due to Computed Tomography on imaging quality has been significantly improved, so commonly widely used in clinical diagnosis. And in the scanning process, CT examination will produce cause harm to human body of ionizing radiation, but reduce CT scanning dose can bring serious impact to the imaging effect. So through continuous research of low-dosed CT denoising, the effective algorithm has important clinical significance and application value.In the process of image noise suppression, removing noise and keeping the detail information is a pair of contradictory conflict. In image denoising the traditional algorithms destroyed the image details, such as texture, edge. This article focuses on the denoising algorithm, which is based on fractional Grünwald-Letnikov(G-L) PMD denoising algorithm for low-dosed CT image, and advances the algorithm based neighborhood variance(NV) adaptive adjustment of fractional order and puts forward the improved image denoising algorithm, which is based on fractional Riemann Liouville(R-L) PMD low-dosed CT, the concrete research content is as follows:Firstly, study the image denoising algorithm based on the fractional-order G-L PMD for low-dosed CT. In view of the low-dosed CT image denoising, comparative experiment is established. By analyzing and comparing, two experimental results are obtained: a. compared to the traditional denoising algorithm, FOPMD algorithm filter noise and keep the details of the image edge and texture. b. Under other conditions are the same, different fractional order to image denoising and keep detail information has important impact: selecting bigger fractional order is conductive to keep the detail information, and selecting smaller fractional order can effectively denoise.Secondly, proposes the denoising algorithm based on the NV adaptive adjustment of fractional order PMD(NV-GL PMD). The idea: select a big fractional order in larger neighborhood variance of texture regional, and select a small fractional order in lesser neighborhood variance of smooth regional, then the neighborhood variance information is used to achieve adaptive control of fractional order. The experimental results show that compared to the fractional order PMD G- L algorithm, the proposed algorithm is improved well in denoising effect of low-dosed CT image and has been effective inkeep image details.Finally, This paper further puts forward an NV-RL PMD denoising algorithm.Because the fractional R-L integral operator can implement keeping edge weak and denoising of noise image to a certain extent, so this article try to put NV-RL PMD algorithm to image denoising. The experimental comparison results show that compared to the NV- GL PMD algorithm, the proposed algorithm has certain enhancement in image denoising and image edge, improve the image visual effect.