Compression of binary images by stroke encoding

The application of digital image systems to document handling in offices is currently limited by the high data transmission and storage costs. This paper describes a new method for encoding binary images of text or line drawings that achieves a high degree of data compaction in return for a small loss of fidelity. Typically, the compressed data stream is only two-thirds of the size of that produced by a reference algorithm employing a 3-pel predictor and a highly sophisticated error encoder. The input image is first processed by a thinning algorithm to extract the centrelines of the strokes that form the characters and lines. A tracking algorithm then connects neighbouring black picture elements into chains whose shapes and positions are encoded to produce the compressed data stream. Outline hardware implementations for the thinning and tracking algorithms are given.