Motion-compensated predictive method for very low bit-rate image coding

Abstract

When the hybrid image coding method using motion-compensated prediction and transform is applied to very low bit-rate (less than 64kbit/s) image coding, it often causes block distortion in decoded images, since motion detection and motion-compensated prediction has been performed based on square block. Recently, as the one of the methods to solve this problem, some motion-compensated predictive methods based on affine transform or bilinear transform were developed. However, these methods can't exactly express the appearance motion in the 2-D projective image plane corresponding to the motion in the 3-D space. In this paper, we propose a new motion-compensated predictive method for very low bit-rate image coding. The proposed method is based on the perspective projection and the gradient of pixel value in time and space domain, and has high ability in motion representation. The computer simulation was performed using moving test images, and the resulting predictive image and SNR were compared with the conventional methods. The results showed that the proposed method has good performance as a motion-compensated predictive method.