Noise Reduction For Low-dose CT Image Reconst-Ruction Technology Base On Projection Domain

In recent years, the technologies of Computed Tomography (CT) have gained rapiddevelopments and were widely used in the fields of clinical medicine, and then people begunto pay attention to the problems in the process of high radiation dose in CT scanning. Due tothe radiation dose has a linear relation with X-ray tube current, it is sure that reduce the tubecurrent can decrease the dose of radiation. But the CT reconstruction may have a seriousdegradation of image quality, then affecting the accuracy of clinical diagnosis. So how toeffectively improve the quality of low dose CT image becomes the focus of CT technologyresearch. Among all of them, in the processing of noise reduction in projection domain databefore CT reconstruction is one of the effective methods. The paper discusses the principles ofCT and the algorithms about reconstruction and filtering, and the main researchful direction isanalyzes the statistical properties and in the processing of the noise reduction in low dose CTprojection data. The summaries of the main contents are as follows:Firstly, the paper describes the backgrounds, purposes and significances of the studies inlow-dose CT, and expounds the current research and developments in this area. Then thispaper elaborates some of the classic image reconstruction algorithm., And gives the imagequality evaluation criteria.Secondly, the main researchful direction of this paper is how to apply the partialdifferential equations to the process of low-dose CT image reconstruction. In view of theshortcomings of existing partial differential equation which do not consider specific featuresof the image in diffusion denoising model, this paper which combined with the inherent characteristics of the image, such as the edge of the image value, variance and the pulse image,proposed an improved image noise reduction algorithm based on partial differential equations.The algorithm used the edge value, variance, and pulses to determine the extents of spread,and thus determined the strength of the anisotropic diffusion, then gained a better effect inreducing the noise of image.Again, combing with non-local means filtering principle, this paper proposes a newpartial differential equations denoising algorithm which based on non-local diffusion andmedian filter. This algorithm added median filter and the variance of the image to traditionalPM model, which better protect the details of the image and limit the spread in the directionwhich parallel to the isolux. The algorithm used non-local average gradient to control thedegrees of diffusion, which better distinguish the edge details of the image and the imagenoise.Finally, this paper presents a new pixel model, fuzzy image reconstruction method offiltering with partial differential equations. The algorithm combines the advantages of thethree methods, image reconstruction can effectively remove the image noise and keep theedges of the image details and contour features.