Microwave Imaging of Nonsparse Domains Using Born Iterative Method With Wavelet Transform and Block Sparse Bayesian Learning

A microwave technique utilizing the combination of wavelet transform, block sparse Bayesian learning (BSBL), and Born iterative method (BIM) is proposed to image nonsparse domains. The wavelet transform is implemented to convert the nonsparse domain into a sparse domain. Then, BSBL framework based on expectation-maximization (EM) algorithm is applied on the BIM model to reconstruct the original profile of the nonsparse domain. The presented imaging results of a nonsparse model indicate that the proposed technique, in comparison with traditional microwave imaging or compressive-sensing (CS) algorithms, achieve very low normalized error rate (NER) at a short computational time using only small number of antennas. The accuracy, robustness, and effectiveness of the proposed method are further assessed by employing it to detect a hemorrhagic brain stroke in a realistic, numerical head model, which is a nonsparse domain. The obtained results indicate the capability for the technique to detect an early stroke in the realistic nonsparse environment of the human head using only six antennas.