Interframe predictive coding of images using motion estimation and vector optimization

A vector quantization (VQ) scheme is proposed for image coding to achieve high compression ratios in visual communication applications. For still images, an adaptive tree structured vector quantization is proposed, and the picture quality of the reconstructed image can be adjusted accordingly by changing the threshold value. For image sequences, motion compensation is employed to reduce the variance of input vectors. Then vector quantizers are designed for prediction errors and utilized to reduce bit rate and to improve the reconstructed image quality. Studies using subband VQ with motion compensation were conducted, and satisfactory results were obtained in terms of implementation complexity and compression ratios.<
>