Adaptive multi-channel prediction for lossless scalable coding

In this paper, a new lossless coding of images is proposed. The method has “scalability”, namely, it can expand a rough image from a part of the bit stream (compressed data) and also expand the original image from the rest. The new method is adaptive to locally varying characteristics such as “edge” in the input image. A new lossless “anti-aliasing” filter is designed to reduce the “aliasing” in the rough image. A lossless multi-channel non-separable 2D filter bank is applied and its filter coefficients are adaptively optimized for each of the pixels according to the local characteristics of the image. It is not necessary to include the coefficients into the bit stream since both of the encoder and the decoder employ the same procedure to calculate the coefficients. Experimental results are indicated to confirm its effectiveness