ABSTRACT
This chapter describes the widely used image Cross Correlation (CC) algorithm in spatial domain and image cross-power spectrum (CPS) algorithm in frequency domain. To meet the real-time requirement of image processing in machine vision applications, a novel Gaussian Muti-scale Fast Registration (GMFR) algorithm is proposed. GMFR avoids the problem of losing details of images in sub-sampling process by means of Gaussian filtering based on muti-scale images. Based on CC and CPS, muti-scale image analysis is introduced. Then GMFR is proposed a performance function is defined to analyze the efficiency of the presented registration algorithm. The performance of GMFR is analyzed quantificationally based on the pre-defined performance function, which shows that the advantage of GMFR algorithm increases exponentially as the image gets larger. The experiment results of lead frame image demonstrate that the GMFR has good robustness when dealing with noised image, and the effciency is improved obviously.
