Construction of Orthonormal Quasi-Shearlets based on quincunx dilation subsampling

Abstract: We consider the construction of orthonormal directional wavelet bases in the multi-resolution analysis (MRA) framework with quincunx dilation downsampling. We show that the Parseval frame property in MRA is equivalent to the identity summation and shift cancellation conditions on M functions, which essentially characterize the scaling (father) function and all directional (mother) wavelets. Based on these two conditions, we further derive sufficient conditions for orthonormal bases and build a family of quasishearlet orthonormal bases, that has the same frequency support as that of the least redundant shearlet system. In addition, we study the limitation of our proposed bases design due to the shift cancellation conditions.