FFT-based 4-parameter global motion estimation

Abstract

Motion estimation and compensation play important roles in video coding. The most commonly used motion estimation technique is block matching. In recent years, the global motion compensation (GMC) is paid great attentions because it is an important tool for a variety of video processing applications including for instance registration, segmentation and video coding. The phase correlation is a typical global motion estimation (GME) method in frequency domain. In this paper, a new 4-parameter GME method is proposed based on the Fourier transform properties of the tow images before and after global motion. At first, the scale and rotation parameters of the affine transform are estimated according to the formulas derived in this paper. Then the effect of scale and rotation of the affine transform are corrected with the estimated parameters. After that, the well-known phase correlation technique is used to determine the two shift parameters. An algorithm according to this principle is proposed in this paper, and simulation results show that the 4 affine parameters can be exactly estimated with our new method.