Research On Image Registration Of Infrared And Visible Image

The visual images has a higher resolution and a better expression of image details and texture information than the infrared images,but the producing of visual image is susceptible to natural environment such as fog weather and lightless condition.Infrared sensor can get images in night and bad weather conditions,but infrared image always has a low resolution.Now more and more visual systems combine this two kind of image sensor to enrich single image sensor's insufficient information.As an important and fundamental method in image processing,image registration is widely used in image mosaic,3D vision,multi-sensor image fusion and other fields.Multi-sensor visual systems also promote the development of image registration technology.In this paper,a feature-based registration method for infrared and visible images is studied.The main work is as follows:1.The principle of SURF feature extraction algorithm is introduced in detail.The whole process from feature point position location to feature point vector descriptor construction is elaborated.The experimental results of extracting infrared and visual image feature points by SURF are given.2.This paper introduces the traditional RANSAC algorithm for eliminating mismatching pairs of feature points,and analyses the main reasons why the method has a poor effect in visible and infrared image registration.To overcome the shortcomings of this method,a new method based on geometric constraint is proposed to obtain correct matching pairs of feature points,and its logical feasibility is analyzed.Finally,the experimental results are given.3.A registration evaluation method based on edge coincidence is proposed.The feasibility of this method is proved.The experimental results show that this method can be applied to the registration evaluation of infrared images and visible images.4.The experimental results of this algorithm on the DSP platform are given.