The Denoising Algorithm Of Low-dose CT Images Based On Shearlet Transformation

Computed Tomography technology has been widely used in clinics of medical and has achieved substantial enhancement both in efficiency and precise. However, the dose of the X-ray radiation is higher and higher within the Sustainable development of CT equipment technology. It is because the quality of CT images with X-ray radiation dose is closely related to and directly proportional relationship. However, the high dose of CT severely limits the scope of application of the CT scan and while people are increasingly concerned about how to reduce the dose of X-ray radiation within ensuring the image quality. In order to improve the quality of low-dose CT images, the article proposes the noise suppression algorithm based on the shearlet transform, according to the images’noise characteristics and the statistical distribution law, and then obtains the CT images for clinical diagnosis.Wavelet transform are used a lot in the digital image processing because of the multi-resolution features, but the wavelet transform cannot be said the graphics performance effectively result of some reasons. For example, it has the less sparseness and can only express appoint singular with the limited direction. In recent years, the shearlet transformation theory is proposed to effectively solve the defects of the wavelet transform. This paper goes into the structural theory and the basic nature of the shearlet transformation, consequently, the threshold denoising algorithm is applied to the low-dose CT images in view of the discrete shearlet transformation. The following is some aspects about the paper.1. This paper analyses the research of the noise characteristics based on the operating principle of the CT scan and the CT image-forming principle, and then observes that the law of the noise in the low-dose CT image is approximation to the Poisson distribution. Accordingly, we establish the noise model.2. This paper has a deep research on the structural theory of the shearlet transform and the nature such as the sparseness, the redundancy, and the time complexity. At the same time, we described the discretization method on the shearlet transformation in detail. On one hand, the shearlet transform can be said the graphics performance effectively and sparsely; on the other hand, it has the multi-scale characteristic, the multi-resolution characteristic and the multi-directional characteristic. For these reasons, it has great feasibility used in the low-dose CT image.3. This paper researches the key issues about the threshold denoising, that is to say, the selection of the basis function, the method of the noise estimation, the selection of the decomposition level and the choice of the threshold function. We emphasize the method of the noise estimation because it affects the calculation of the subsequent threshold function. However, the ability of the Ascombe transform is to transform the random variables submitted the Poisson distribution into alternative random variables which is submitted to the Gaussian distribution. Thereby, this method reduces the error caused by the directly noise estimation about the Gaussian distribution.4. As is well-known, different basis functions with different time-frequency characteristics, therefore, it will produce different result on the denoising of the CT image. This paper analyses the effect on the denosing of the low-dose CT image using the different basis function. At the same time, the experimental simulation demonstrates the best basis function and the decomposition level on the shearlet transformation.