Binary shape coding using finite automata

Two new binary shape-coding techniques that are based on finite automata methods are proposed. Both contour-based and bitmap-based approaches to shape coding are investigated using one-dimensional weighted finite automata (1D-WFA) and generalised finite automata (GFA) algorithms, respectively. We evaluate the fidelity of shape representation, using 1D-WFA and GFA methods in intraframe mode and compare the performance of the proposed coding techniques against each other and with the benchmark, context-based arithmetic encoding (CAE) of shapes used in MPEG-4. It is found that the GFA method is more suitable for video applications than 1D-WFA and therefore it is adapted to operate in interframe mode. More importantly, GFA is the first shape-coding technique reported to date that has the unique advantage of shape processing in the compressed domain. This is due to the fact that the shape representation in the compressed domain using GFA facilitates processing at the expense of less compression efficiency compared with MPEG-4 CAE. Moreover, shapes encoded using the GFA method can be decoded at any desirable resolution