magnetic resonance imaging (mri) is a notable medical imaging technique that makes of phenomenon of nuclear magnetic resonance. because of the resolution and the technology being harmless, mri has considered as the most desirable imaging technique in clinical applications. the visual quality of mri plays an important role in accuracy of medical delineations that can be seriously degraded by existing noise during acquisition process. therefore, denoising is of great interest for diagnostic aims and also the successive throughput of automatic computerized analysis such as segmentation and classification. noise in magnitude mri is usually modeled by rician distribution which introduces a signal-dependent bias and reduces the image contrast. in this thesis an efficient approach for enhancement of the noisy magnitude mri based on the linear minimum mean square error (lmmse) estimator is introduced. the natural redundancy of the acquired mr data is employed to improve the performance of unknown signal estimation. since in practice, the mr data is in a large majority 3-d, the proposed method is developed to deal with 3-d mr volumes as well as 2-d mr images. the quantitative and qualitative metrics have been used to demonstrate and compare the performance of the introduced approach with several state-of-arts denoising schemes. experimental results show the suitable denoising ability of the proposed method. moreover, preservation and restoration of delicate structural details is a critical task in medical image analysis. the relative experiments show the superior performance of the proposed method in comparison with other denoising schemes in the case of restoration of delicate features.