Multi Scale Image Fusion through Laplacian Pyramid and Deep Learning on Thermal Images
Main Article Content
Abstract
Scale is an imperative parameter in image processing. Images can contain different objects and regions with variations in size as well as variations in resolutions. Appearance of objects in the image mainly depends on the scaling parameter. In order to analyse the images in a better way it is necessary to represent images in different scales because a single scale image is restricted to a fixed bounded window. To have useful information from the available images there is a necessity of representing images in a multi scale fashion. Thermal imaging means producing visible images from invisible thermal radiation that is it is a method of generating images with the help of heat given to the object. Development of thermal images took place in 1950s and 60s for the purpose of military applications. In this paper, focussed in providing advanced and enhanced fusion on thermal images by using multi scale laplacian pyramid in combination with deep convolutional neural networks. The proposed approach includes fusion method based on laplacian pyramid and deep learning on thermal images for clinical applications, The results of this method shown improved results in identifying abnormalities in the medical images with the help of seven different image metrics.. In the fields of defence and law enforcement object or person identification is needed for their work. The resultant fused image of the second method gave better image metric values in order to provide more significant visual information.