FINITE-STATE RESIDUAL VECTOR QUANTIZATION USING A TREE-STRUCTURED COMPETITIVE NEURAL-NETWORK

Finite-state vector quantization (FSVQ) is known to give better perfor mance than the memoryless vector quantization (VQ). This paper present s a new FSVQ scheme, called finite-state residual vector quantization (FSRVQ), in which each state uses a residual vector quantizer (RVQ) to encode the input vector, This scheme differs from the conventional FS VQ in that the state-RVQ codebooks encode the residual vectors instead of the original vectors, A neural network predictor estimates the cur rent block based on the four previously encoded blocks, The predicted vector is then used to identify the current state as well as to genera te a residual vector (the difference between the current vector and th e predicted vector), This residual vector is encoded using the current state-RVQ codebooks, A major task in designing our proposed FSRVQ is the joint optimization of the next-state codebook and the state-RVQ co debooks, This is achieved by introducing a novel tree-structured compe titive neural network in which the first layer implements the next-sta te function, and each branch of the tree implements the corresponding state-RVQ, A joint training algorithm is also developed that mutually optimizes the next-state and the state-RVQ codebooks for the proposed FSRVQ, Joint optimization of the next-state function and the state-RVQ codebooks eliminates a large number of redundant states in the conven tional FSVQ design; consequently, the memory requirements are substant ially reduced in the proposed FSRVQ scheme, The proposed FSRVQ can be designed for high bit rates due to its very low memory requirements an d the low search complexity of the state-RVQ's, Simulation results sho w that the proposed FSRVQ scheme outperforms conventional FSVQ schemes both in terms of memory requirements and the visual quality of the re constructed image. The proposed FSRVQ scheme also outperforms JPEG (th e current standard for still image compression) at low bit rates.