Infrared and visible images registration using BEMD and MI

Enhanced vision especially in poor weather conditions can be achieved by image fusion between the infrared and visible images. Image registration is a crucial step in image fusion. In order to register the infrared and visible images captured in poor weather conditions, the registration method combing the Bidimensional Empirical Model Decomposition (BEMD) with mutual information is proposed in this paper. After the infrared and visible images are decomposed into a set of Intrinsic Mode Functions (IMFs) and residue respectively by BEMD, the parameters of affine transform can be obtained through using the maximum mutual information registration between IMFs images. Then, the transform matrix defined by these parameters is used to the source infrared image and the final registered image can be obtained through interpolation and resampling for the transformed image. This method is used to register two video sequences captured in dense fog. Experimental results show that this method gives better performance compared to other two methods.