Infrared and Visible Light Image Fusion Method for Three-Channel Inseparable Wavelet and Deep Learning
At present,with the increasing maturity of the basic discipline of computer science,deep learning and artificial intelligence are also developing rapidly,deep learning can extract deep features of images,which can effectively make up for the deficiency of deep information extracted by artificial filters.The paper puts forward the infrared and visible image fusion algorithm which is two-dimensional with different channels for inseparable wavelet and deep learning.The paper introduces the theoretical basis of three-channel inseparable wavelet,constructs the corresponding filter set,and decomposes the image to obtain low frequency subimage and high frequency subimage.The low-frequency images are fused by taking the maximum of the regional energy for reuse;the VGG-19 network trained by ImageNet was used to extract more and deeper feature information from the high-frequency subimages.The weight diagram is calculated by softmax operator of the image information of each layer,and the weight diagram is used to obtain the high-frequency subimage of the final stage.The same approach was used to manipulate the first three layers of the network,and the three high-frequency subgraphs were fused using the maximum value selection strategy to obtain the final high-frequency subimages.The inseparable wavelet inverse transformation is completed to obtain the final fusion image.Compared with other related methods,it has a better effect performance in the subjective and objective aspects.
万方数据
维普
