Nearly-lossless autoregressive image compression

A flexible accurate technique for improving image compression properties with linear image compressing transforms is presented. This is a reliable iterative algorithm for the nearly-lossless autoregressive (NLAR) compression of images, which permits control of errors produced in each pixel intensity while preserving the original image intensity range. The convergence of this new algorithm is demonstrated, both theoretically and numerically. Its relationship to other types of nearly-lossless compression, as well as to traditional lossless and lossy approaches, is discussed.