Improved compression of MRSI images involving the discrete wavelet transform and an integrated two level restoration methodology comparing different textural and optimization schemes

Abstract

This paper suggests a novel MRSI image compression scheme, using the discrete wavelet transformation (DWT) and an improved integrated Bayesian reconstruction approach involving a parameter independent optimization scheme. The suggested methodology is based on maintenance of important second and higher order correlation features of DWT coefficients and image pixel intensities. While adversary image compression methodologies utilizing the DWT apply it to the whole original image uniformly, the herein presented novel approach, extending previous attempts of the same author, involves a refined DWT compression scheme. That is, different compression ratios are applied to the detailed wavelet coefficients belonging in the major regions of interest, clustered by employing textural descriptors as criteria in the image or transform domain, integrating different textural methods. Restoration of the original MRSI image from its corresponding regions of interest compressed images involves the inverse DWT and a sophisticated two stage Bayesian restoration approach, not requiring any user defined parameters, comparing conjugate gradient and Genetic algorithm optimization processes involving a refined objective function. An experimental study is conducted to qualitatively assessing the proposed schemes in comparison with the original DWT compression technique as well as with other rival approaches based on DWT, when applied to a set of brain MRSI images.