Bi-directional Intra Prediction Based Measurement Coding and Encoder Rate Control for Compressive Sensing Images

This work presents a bi-directional intra prediction-based measurement coding algorithm to reduce spatial redundancies found in block-based compressive sensing schemes. By exploiting the fixed basis of a structured measurement matrix that allows members in measurement to be traceable to the position of the pixel, we choose the closest element between neighboring to make the prediction candidates. Following that, we apply the triangle method to scalar quantization, resulting in a controllable encoding bit rate that maximizes performance per block against bit rate consumption. Experiment results show that this work improves PSNR by 0.01 - 0.02 dB and bit rate reductions by 19% on average and up to 36% compared to the state-of-the-art.