Motion Adaptive Wavelet Thresholding for Recovery of Compressively Sampled Static and Dynamic MR Images

Abstract

Iterative shrinkage algorithms like parallel coordinate descent and separable surrogate functional use wavelet thresholding with uniform and empirically selected threshold values to recover the under-sampled magnetic resonance (MR) images. In this paper, an adaptive thresholding parameter, for the recovery of static and dynamic MR images, is derived and used in wavelet domain shrinkage. A modified iterative shrinkage thresholding algorithm based on the derived parameter is also proposed. Simulation results show that adaptive wavelet thresholding yields significantly higher signal-to-noise ratio and correlation than the fixed thresholding value. The algorithm based on the adaptive threshold is experimentally tested for static and dynamic MR images with varying acceleration rates, and it has been shown that it outperforms the fixed thresholding value algorithm.