Directional Analytic Discrete Cosine Frames

Block frames called directional analytic discrete cosine frames (DADCFs) are proposed for sparse image representation. In contrast to conventional overlapped frames, the DADCFs require a reduced amount of 1) computational complexity, 2) memory usage, and 3) global access. These characteristics strongly required current high-resolution processing. Specifically, we propose two based on transform (DCT) sine (DST). The first DADCF is constructed from parallel separable transforms DCT DST, where DST permuted by row. second also designed while customized have no DC leakage property which desirable Both rich directional selectivity with slightly different each other they can be implemented as non-overlapping block-based transforms, especially form Parseval block low redundancy. We perform experiments evaluate our compare them in recovery.