Technique for image fusion based on nonsubsampled shearlet transform and improved pulse-coupled neural network

Abstract

A new technique for image fusion based on nonsubsampled shearlet transform (NSST) and improved pulse-coupled neural network (PCNN) is proposed. NSST, as a novel multiscale geometric analysis tool, can be optimally efficient in representing images and capturing the geometric features of multidimensional data. As a result, NSST is introduced into the area of image fusion to complete the decompositions of source images in any scale and any direction. Then the basic PCNN model is improved to be improved PCNN (IPCNN), which is more concise and more effective. IPCNN adopts the contrast of each pixel in images as the linking strength β , and the time matrix T of subimages can be obtained via the synchronous pulse-burst property. By using IPCNN, the fused subimages can be achieved. Finally, the final fused image can be obtained by using inverse NSST. The numerical experiments demonstrate that the new technique presented in this paper is competitive in the field of image fusion in terms of both fusion performance and computational efficiency.