Research On Fusion Method Of Infrared And Visible Images

Multi-source image fusion is a hot research direction in recent years,involving sensor technology,digital image processing,computer vision,and artificial intelligence.Infrared and visible light images have important application significance in many fields such as social security and video surveillance due to their good complementary nature.The research in this paper will be based on the fusion of infrared and visible light.The main research results are as follows:(1)An adaptive PCNN image fusion algorithm based on NSST domain is proposed.Firstly,the source image is decomposed by NSST to obtain high frequency coefficient and low frequency coefficient.Compared with other multi-scale decomposition methods,NSST is an optimal and sparse function representation method.Secondly,it is designed for the nature of image high and low frequency information.Different fusion criteria are used to apply the regional energy and fusion method to the low-frequency coefficients,and the high-frequency coefficients are combined by the QPSO-PCNN algorithm.Finally,the fusion coefficients are reconstructed by the NSST inverse transform to obtain the fused images.(2)An image fusion algorithm based on convolutional neural network is proposed.By studying the properties of infrared and visible images,by designing the similarity threshold analogy multi-focus image fusion,the 2-channel network is designed to obtain the weight map to fuse the similarity points above the threshold,and the local energy fusion method is designed to similarity.Convergence is performed at points below the threshold.Considering the human visual system and using pyramid decomposition,the decomposition coefficients of each layer are fused at various resolutions to further improve the fusion quality.The experimental results show that the two algorithms proposed in this paper have obvious superiority,and the image quality is greatly improved in both subjective observation and objective data.