Termination Factor for Iterative Noise Reduction in MRI Images Using Histograms of Second-order Derivatives.

Histograms of second-order derivatives are generated from the pixel data of MRI images. The histograms are then used to calculate a factor that is to be used for iterative processing. The factor is intended to limit the number of iterations, with the goal of preventing further loss of detail. The factor uses two conditions that depend on the profiles of the histograms. The methodology uses sample MRI images and versions of these images with Rician noise introduced into them. The noisy images are subjected to iterative noise reduction with a recursive averaging filter. The control tests in the methodology use the ground truth images to limit the number of iterations, with PSNR and SSIM peaks used as the measurements for determining when the iterations stop. The other tests use the proposed termination factor for the limitation. The results of the tests are compared to determine the effectiveness of the termination factor. The proposed termination factor does not cause divergence, but there are still different numbers of iterations in the case of MRI images with image subjects that have discrete regions and details resembling noise. The tests also reveal that differences between the histograms of derivatives and Laplace curves have to be retained in order to prevent loss of information. [ABSTRACT FROM AUTHOR]