Multi-scale image compressed sensing with optimized transmission

Existing works in image compressed sensing (CS) often focus on improving the rate-distortion sensing performance but have less consideration of the effect of channel errors on the transmission of CS measurements. This paper explores how the transmission channel errors affect the PSNR performance of the quantized sensing measurements and then increases the resistance of the transmitted data to the noisy channel. We show that the multi-scale block based compressed sensing (MSBCS) using quantization with differential phase code modulation (DPCM), though achieves compression efficiency higher than the regular scalar quantization-based CS, is more vulnerable to channel errors. To improve the error resistance, a method of optimized energy allocation across bit-layers in quantized measurements is presented, which leads to a large PSNR performance improvement for CS data transmission in wireless environment.