MR image de-noising algorithm based on improved weickert structure tensor model

In allusion to P-M do not take the directional construction of local image into consideration, and to solve the false edge caused by using Weickert diffusion tensor to process the flat part of the image. A MR image de-noising algorithm based on improved Weickert diffusion tensor model has been proposed in this paper, by using continuous threshold, selecting appropriate eigenvalue modification method timely, and combining edge enhancement and coherent enhancement's eigenvalue together. Experiments are implemented on Lena image and brain MR images, and the average index of MSE, SNR, PSNR under different noise levels have been gotten. Compared to P-M model, EDD and CED, our algorithm achieved the best result. The experiment results showed that the algorithm can effectively remove noises from MR image.