Low-data-rate visual communication using cartoons: a comparison of data compression techniques

An analysis of two-level picture compression techniques applied to low-resolution moving images is reported. The object is to discover the most suitable technique for visual communication at low data rates using feature-extracted `cartoons¿¿. Reversible facsimile compression techniques are reviewed, then results presented to demonstrate their relative performance. In general, three-dimensional techniques prove to be most efficient, since they exploit both spatial and temporal relationships in the picture. However, the lower compression of two-dimensional coding is balanced by its superior error recovery performance, and it is therefore recommended for very low-data-rate transmision over conventional telephone lines, where error rates are high. From the tested range of 2-D and 3-D techniques, `relative address coding¿¿ is most successful, being about 10% more efficient than its nearest rival, except for high-resolution pictures with little movement, where `block location coding¿¿ of diffence frames shows a small advantage. Application of irreversible preprocessing improves the compression performance, but the gain is small, and requires additional processing power.