Muliresolution, Adaptive Vector Quantization and Perceptual Based Multiview Image Codec

This paper presents a multiresolution adaptive vector quantization and perceptual based multiview image coding scheme. It decorrelates the input views into a number of subbands using a lifting based wavelet transform. The coefficients in the same subbands of different views are divided into vectors and then joined together. The resulting vectors are then vector quantized using an adaptive vector quantization scheme. Perceptual weights are designed for different viewing distances and used in the vector selection and bit allocation stages of the adaptive vector quantization technique. In order to evaluate the performance of the proposed codec, two sets of multiview test images were coded using the proposed codec with and without employing perceptual weights and the monoview vector quantization coding algorithm. Results indicated that the proposed codec with and without using perceptual weights significantly outperform the basic vector quantization technique. Results also showed that the proposed technique with perceptual weights gave superior objective and subjective image quality compared to the algorithm without perceptual weights.