Signal denoising is the process of removing noise signals from a noisy image. But still, it remains as an important issue for the biomedical engineering. Due to the highly controlled imaging environment, the imaging process often creates noise, which seriously affects the analysis of the medical image. To solve the issues of the denoising in biomedical engineering efficient learning algorithms are used, for sparse representations of the data. Sparse representations are representations that account for most or all information of a signal with linear combination of small number elementary signals. At present algorithm such as Dictionary learning algorithm is exploited for spare representation of the data and this algorithm can be applied to 3 D Medical Image Denoising. The learning approach is involves two main parts: sparse coding and dictionary updating. Denoising of 3-D medical image that involves large number of slices is to denoise each single slice using separately learned dictionaries. In 3-D medical image denoising fixed square patches are applied into denoising method. This leads to computational complexity and also no good approximation of image can be constructed. So instead of adopting the fixed square patches, shape-adaptive patches can be applied as in the SA-DCT into the denoising method. Furthermore SA-DCT algorithm is proposed for image filtering. Such adaptation strategy enables accurate preservation and reconstruction of image details and structures and yields estimates with a very good visual quality. © 2013 IEEE.