Directional block compressed sensing for image coding

Lei Liu, Anhong Wang, Kongfen Zhu, Chunyu Lin, Yao Zhao

Abstract: Compared with traditional Nyquist sampling, compressed sensing (CS) enables a highly precise reconstruction of the signal from fewer measurements, suggesting great potential for efficient and simplistic data acquisition. In this paper, we propose a directional block-based compressed sensing (DBCS) scheme for image coding, where the directionalities inherently exhibited within image blocks are exploited as the “a priori” information. The image block is first directionally scanned following the dominating direction of its edges/textures. Then the vectorized image block is sampled by a block-based compressed sensing (BCS) method. At the decoder, each image block is recovered and then rearranged by the corresponding inverse-scan to obtain the recovered image. Experimental results show that the proposed DBCS scheme outperforms BCS due to the exploitation of the directional information within image blocks.