Differentiation-based Multi-resolution Approach for Lossless Image Compression

We design a differentiation-based multi-resolution approach for lossless image compression. The differentiation technique uses six appropriately weighted adjacent pixels (i.e., the pixels located at the west, west-west, north-west, north, north-east, and north-north of the predicted pixel) to estimate the predicted pixel intensity. This differentiation technique can also be considered as an additional rule for the seven predictors used by the lossless JPEG. The prediction error is decomposed by a (2, 2) integer wavelet transform [1] (i.e., the number of vanishing moments of the analyzing high-pass filter and the synthesizing high-pass filter is 2 and 2 respectively) to improve the prediction. The reasons to choose the (2, 2) integer wavelet transform follow [2]: • It maintains the highly important linear phase constraint by relaxing the orthogonality constraint and associating with short length filters. Therefore, it is a suitable tool to compress images regarding minimization of computational time and maximization of compression ratios.