Encoding-interleaved hierarchical interpolation for lossless image compression

In this correspondence an improved version of the Hierarchical INTerpolation (HINT) algorithm is proposed for multi-resolution reversible compression of still images. The efficiency of the basic scheme is increased by splitting the non-separable interpolation process into two cascaded directional steps interleaved with encoding. Interpolation errors from the former are encoded to yield error-free pixel values, from which the left-over pixels are interpolated in the latter step. The outcome benefit is twofold: interpolation is executed starting from all error-free pixel values, and different directional correlations may be exploited. In addition, the differences between true and interpolated pixel values are encoded cyclically with respect to the range of pixel values within square blocks, thus saving extra bits compared to plain entropy coding. Results and comparisons witness the improved coding performance.