Research On Image Registration Based On ELM And Demons Algorithm

Image registration is of important application in image processing.In recent years,with the development of image processing technology,image registration has been widely used.Image registration is divided into two branches,which are gray-based image registration and feature-based image registration.These two registration techniques are used in longitudinal tracking,diagnosis of diseases in medicine,panoramic stitching of images,processing of satellite images,and intelligent machines.Image registration is designed to calculate two images or multiple images correspondence of location coordinates in space coordinates.Image registration can also be called spatial regularization,image alignment.The diversity of processing methods,establishment of image registration model and the corresponding optimization reflect the meaningful of this area.The process of image registration can be regarded as the learning process of neural network in machine learning.Based on this idea,we combined extreme learning machine with famous Demons algorithm to realize image registration.The main research results are as follows:By combining extreme learning machine with famous Demons algorithm,we proposed a new image registration algorithm.The proposed algorithm uses extreme learning machine to obtain registration information and improve Demons algorithm.The remote sensing image registration method is studied.Based on extreme learning machine and registration technology a new method is proposed which is applied to remote sensing image registration.In the proposed algorithms,the image registration process does not solely depend on the gradient information of the image,so it avoids the mutual interference between the points.The experiment proved that the proposed algorithm is effective and feasible.