Regional sub-block matrices based multiple regularization and biomedical image reconstruction

Abstract

A regional information based multiple regularization is studied for solving biological inverse problem. Inverse problems are usually optimized and solved by Newtons method and its variants. Optimization based on calculus is extremely localized. The regional gradient in calculus is more important for local physiological changes. A sub-block based multiple regularization is proposed in Gauss-Newtons method for biological diffuse optical tomograph (DOT). A study of single step regularization (STR) method and proposed subblock based multiple regularization method has been carried out. The reconstructed image analysis shows a significant improvement in the proposed method.