CS image reconstruction by novel small length LDPC code

Abstract

Compressed sensing (CS) is capable of reconstructing the data signal from very small number of measurements. However the reconstructed image quality gets highly affected in the case of far end reconstruction involving transmission through a channel. In this work, a novel small length binary low density parity check (LDPC) code for far-end transmission of compressively sampled signals is proposed. Smaller length of LDPC code means faster computation. Also the code rate of proposed LDPC code is constructed to be 0.6 as in an earlier work by the same group it has been demonstrated that quasi-cyclic LDPC codes, with code rate 0.67 gives optimum performance with CS. The proposed code has a base matrix of size 4 × 10. The length of the LDPC code constructed, can be adjusted by a base matrix multiplying factor. The multiplying factor varies from 10 to 30 in steps of 4. The constructed code has a variable length between 100 and 300, in steps of 40. Simulation results of the proposed LDPC code show good performance when decoded iteratively by min-sum decoding algorithm. The proposed LDPC code also gives improved CS reconstruction for far end reconstruction.