Research On Denoising And Interpolation For Medical Image

As an auxiliary method in clinical diagnosis, the processing of medical image signal plays a much more obvious role with the rapid development of image processing technology. The increasing demands for clinical application requests higher quality of medical images than before. High quality medical images are not only the crucial diagnosis references for doctors, but also provide reliable pre-process results for feature extraction and classification. At the sampling stage, the output signal quality would always be restricted by some condition limitations, which make the resolution lower than expected.Besides that, noise signal could also become another reason for image degradation when at the encoding stage and the transition stage. At the very beginning of this paper, image prior models would be introduced as theory basis for denoising and interpolation. This paper focus on some popular models including group sparse representation model, local autoregressive model and non-local similarity model and analyze how these models could be used in medical image processing. For the purpose of effective denoising for medical images, this paper improves the traditional group sparse representation model by adaptive threshold and group block numbers to enhance the denoising performance. In traditional interpolation methods, only one prior model would be employed which did not archive good results. In this paper, we present the regularization framework ofmedical image interpolation combined with the proposed soft data accuracy. The experimental results show that our proposed denoising method which based on adaptive block number within the group and weighted soft-threshold achieves better results than traditional methods.Meanwhile, our proposed interpolation method also performed better than traditional algorithms.