Wavelet-based compressive sensing for head imaging

A wavelet based compressive sensing technique for head imaging is presented. The non-sparsity of the dielectric profile of the human head brings about difficulties when applying traditional compressive sensing technique to image the profile of the head. In this paper, the wavelet transform is implemented to convert the non-sparse profile into a sparse domain then a compressive sensing framework named block sparse Bayesian learning (BSBL) is applied on the Born iterative method (BIM) model to reconstruct the original profile of the non-sparse domain. The proposed method is evaluated on a realistic human head phantom. The results show that a very low normalized error rate at a short computation time using small number of antennas can be achieved. The obtained results indicate that the presented technique can enable detecting an early stroke in the realistic non-sparse environment of the human head using only six antennas.