Enhancement of infrared-based image identification system for security robots by image decomposition

Abstract

In this paper, a new infrared-based face identification system based on an input image decomposition is presented. The proposed image decomposition is a novel approach to transform an infrared image into a multilayer infrared image based on energy levels. Image decomposition is done by firstly applying false-color palette to the input image which will then be decomposed according to each of the false-color layer. The proposed approach is able to identify correct images from manifold poses by using frontal view of the infrared images and its decomposition layers as training images. In this work, the proposed method was verified through the implementation on a face recognition system, which was based on wavelet and neural network classification. The application of this finding is substantial, for security and rescue robots, for example, where a single image is acquired and should be critically identified.