The skip-innovation model for sparse images

On sparse images, contiguous runs of identical symbols often occur in the same coding context. This paper proposes a model for efficiently encoding such runs in a two-dimensional setting. Because it is model based, the method can be used with any coding scheme. An experimental coder using the model compresses CCITT facsimile documents 2% better than JBIG and is more than three times as fast. A low complexity application of the model is shown to dramatically improve the compression performance of JPEG-LS on structured material.