An Efficient coding algorithm for affine motion parameters

Abstract

We describe an efficient method for the coding of the affine motion parameters. Conventional direct quantization method consists of a linear quantizer and an entropy coder. Although this is the easiest way to code affine parameters, it has some problems. First, its implementation requires too many quantizers and entropy code tables. Second, the direct quantization method does not take into account the interrelationship between the parameters. Therefore, it shows much degradation after motion compensation if the bit rates allocated to the parameters for the transmission or the storage are reduced. To overcome these problems, we propose a new method that deduces the real translational motion from the six affine motion parameters, which have more physical meaning and a strong temporal correlation. In addition, the scaling and rotational motion are transformed into 2-D vector elements, and then the vector quantization method is used to code them. Simulation results show that our proposed scheme provides better performance than the conventional direct quantization method. (C) 2001 Society of Photo-Optical Instrumentation Engineers.