A new method for coding scanned halftone images is proposed. It is inf
ormation-lossy, but still preserving the image quality, compression ra
tes of 16-35 have been achieved for a typical test image scanned on a
high resolution scanner. The bi-level halftone images are filtered, in
phase with the halftone grid, and converted to a gray level represent
ation. A new digital description of (halftone) grids has been develope
d for this purpose. The gray level values are coded according to a sch
eme based on states derived from a segmentation of gray values. To ena
ble real-time processing Of high resolution scanner output, the coding
has been parallelized and implemented on a transputer system. For com
parison, the test image was coded using existing (lossless) methods gi
ving compression rates 2-7. The best o these, a combination of predict
ive and binary arithmetic coding was modified and optimized achieving
a compression rate of 9.