VQ-Based Masks for Transform Coding

Abstract

In this paper, an approach based on VQ to find masks for transform image coding is presented. In the approach, the key role of VQ is to find representative image blocks and use them to obtain masks for different classes of image blocks. Then an adaptive image coding system is proposed where VQ-based masks are employed to select significant transform coefficients. Two examples are provided to investigate the coding performance of proposed system. In the simulation, a self-organizing mapping (SOM) neural network, which is considered as neural VQ, is used because of its good topology-preserving property. Simulation results indicate that the proposed coding system is feasible and promising.