Research On The Denoising Algorithm Of Medical X-CT Image Based On Contourlet Transform

Medial images processing has been a heated topic in medical research field. Medicalimages process, if not performed properly, will bring a negative impact on the quality ofmedial images, which will direactly affect doctors to make proper diagnosises and adoptthe best possible treatments for the patients. CT image is widely applied in madicaldiagnoses and researched owing to its such advantages as high density resolution, speed,clear image, explicit anatomy, as well as its ability to settle on qualitative diagnose.However, the noise is contained in CT image, image fuzzy, increasing the difficulty of thedoctor to the diagnosis, the effect of clinical application, so the CT image de-noisingmethod, retain more image details, has the important practical significance.Firstly, this thesis respectively introduces the advantages and disadvantages of CT,MRT and supersonic diagnostic set and their theories of image formation. Then thedevelopment of CT machine is also introduced, mainly the researches on denoisingAlgorithm, when Contourlet convertion is adopted. The main purposes of this article areas follows:(1) to introduce traditional denoising methods for CT images: median filtering,wiener filtering, wavelet transform, morphological filtering and their relevant applications.(2) to introduce the development and application of Contourlet convertion. Besides, italso includes that how it can be applied to denoise the CT images. In order to iliminateGaussian noise and improve the quality of CT images, some methods are conbinedtogether, such as morphological filtering, wiener filtering,and Contourlet convertion.(3)In such situation that Gaussian noise and impulsive noise co-exist in the CT images, CT images can be dinoised by combining modified adaptive median filtering and Contourlet convertion, based on the facts that median filtering can filter out impulsive noise andthat Contourlet convertion can deduce Gaussian noise. The result shows that the process has remained more image details and has effectively strained Gaussian noise and impulsivenoise.