Rate-Distortion Optimized Vector Spiht Forwavelet Image Coding

In this paper, a novel image coding scheme using rate-distortion optimized vector quantization of wavelet coefficients is presented. A vector set partitioning algorithm is used to locate significant wavelet vectors which are classified into a number of classes based on their energies, thus reducing the complexity of the vector quantization. The set partitioning bits are reused to indicate the vector classification indices to save the bits for coding of the classification overhead. A set of codebooks with different sizes is designed for each class of vectors, and a Lagrangian optimization algorithm is employed to select an optimal codebook for each vector. The proposed coding scheme is capable of trading off between the number of bits used to code each vector and the corresponding distortion. Experimental results show that our proposed method outperforms other zerotree-structured embedded wavelet coding schemes such as SPIHT and SFQ, and is competitive with JPEG2000.