A multi-scan adaptive linear prediction approach for lossless image compression

This paper presents a block based lossless compression approach for gray scale images. Multiple scanning methods are applied to each block, and a newly proposed adaptive linear prediction is performed. There are different prediction residuals obtained corresponding to different context based on multiple scanning. We choose the best residual for coding. That is, rather than relying on any single scanning, our approach is to select a scanning produces the best result on each block. The prediction coefficients are updated during the scanning to optimize the coding accuracy. Experiment results show that our method outperformed JPEG-LS 4-5% in compression efficiency.