A Low Overhead Image Registration Algorithm using DWT and WIPSO for Resource Constrained SBC based Embedded System Application

Abstract

Image registration is a vital but integral component of any image processing application. Fundamentally, image registration involves transforming two or more sets of image data derived from different sources, collected at different instant of times, taken from different perspectives and from different viewpoints, all from different frames of reference into an image in a single coordinate system, a single frame of reference. Over the years, there have been many image registration techniques developed and used in variety of application fields. This paper proposes a computationally low overhead image registration technique that has been tested with variety of images using weight improved particle swarm optimization (WIPSO) algorithm integrated with discrete wavelet transform (DWT). Fundamentally, the proposed scheme comprises of a two-step approach, where the first step involves extraction of the random image from the source image which will serve as an ingredient for the formation of the particle swarm to be used in the WIPSO and the second step involves an explorative search for the target image in the area of interest from the selected population using WIPSO technique. Extensive simulation has established the effectiveness of the proposed soft-computing technique for variety of image registration application; even suitable for deployment in many resource-constrained single board computer based embedded system applications.