A Low Complexity Algorithm for Global Motion Parameter Estimation Targeting Hardware Implementation

Abstract

Now-a-days image alignment is one of the most widely used techniques in computer vision. Image alignment has many applications in fields as diverse as video surveillance, computer vision, medical imaging, and video coding. The estimation of an objects’ motion is a key step in image alignment. In this paper, we will present a low-complexity algorithm for estimation of motion parameters. Most of the motion parameters estimation algorithms are carried out with a precision of 8 bits per pixel; here we propose an algorithm using only 1 bit per pixel resulting in lower complexity. The proposed method includes a technique for calculating the gradient of the sum-of-squared difference (SSD) using XOR operations instead of multiplication. Experimental results show that the proposed method compares favorably with registration using the full precision of the input images.