Block Based Compressive Sensing for GPR Images by Using Noiselet and Haar Wavelet

ompressive sensing (CS) is a new method for image sampling in contrast with well-known Nyquist sampling theorem. In addition to the sampling and sparse domain which play an important role in perfect signal recovery on CS framework, the recovery algorithm which has been used also has effects on the reconstructed image. In this paper, the performance of four recovery algorithms are compared according to visual evaluation and image assessment parameters where noiselet and Gaussian used as the sampling domain and Fourier transform (FT), discrete Cosine transform (DCT) and Haar wavelet transform (WT) used as the sparse domain. Furthermore, the same sampling and sparse domains are also used for ground penetrating radar (GPR) raw data image. Due to the big size of GPR image and high computational expenses, the block-based adaptive sampling based on edge detection is used.