Infrared and visible light image fusion algorithm based on FCM and guided filter

Abstract

In order to better extract the infrared target information of images in dark scenes and retain more background texture details, an infrared and visible light image fusion algorithm based on fuzzy C-means clustering (FCM) and guided filter is proposed. Firstly, the target information is extracted from the source infrared image by FCM, and the target area and background area of the infrared image are obtained. Then, the target region coefficients and background region coefficients are decomposed into their respective high-frequency and low-frequency subband coefficients by using non-subsampled shearlet transform (NSST). Then, according to the different characteristics of different regions, different fusion strategies are adopted. In order to retain more target information, low-frequency subband coefficients of infrared image target area are selected as fusion coefficients of low-frequency target area, and high-frequency subband coefficients of infrared image target area are selected as fusion coefficients of high-frequency target area. In order to keep more texture details, the method of maximizing low-frequency subband image coefficients and information entropy is adopted in the fusion of low-frequency background region. The method of guided filter combined with dual-channel spiking cortical model (DCSCM) is used in the fusion of low-frequency background region. Finally, the final fusion image is obtained by NSST inverse transform. Simulation results show that compared with the existing algorithms, the fusion image obtained by this algorithm has prominent infrared target in subjective vision, clear background texture details and high hierarchy. In objective evaluation, the indexes are better than other algorithms as a whole.