Subspace imaging compressive sensing

A new compressed image sensing approach is presented. The approach departs from conventional sensing mechanism which seeks incoherency between the sensing and representation vectors. The subspace where most energy of the image lies in is first identified (estimated). Sensing vectors are then selected in the subspace. In doing so, base vectors of discrete cosine transform are used as representation vectors, and low-frequency members of the base vectors are considered to form the subspace. Of those selected base vectors some are used as sensing vectors which are phase shifted to enhance incoherency. Experimental results prove that the new approach is significantly better than random sensing as previously used for compressed sensing.