Infrared And Visible Image Fusion Based On Oversampled Graph Filter Banks

Infrared and visible image fusion is an important branch in the field of multi-sensor image fusion,which can effectively integrate the key information of infrared image and visible image.A single visible image is susceptible to the influence of weather and illumination intensity,but the visible image has a strong contrast and can reflect more detailed information.And the contrast of infrared image is low.Even in the case of low visibility,the infrared image can easily find targets.Therefore,the integration of the information in the two images is helpful to enhance and discover targets.Multiscale transform is the mainstream method for infrared and visible image fusion.Some multiscale transform methods include wavelet transform,pyramid transform and so on.These methods have the advantages of providing sensitive sharp contrast change information of human visual cortex and preserve the details of different source images.However,these methods do not take full account of the spatial consistency of source images and cannot be preserved more texture and edge information of the source images.Therefore,this paper proposes a new method for image fusion-oversampled graph filter banks,which has more sophisticated spectral decomposition capabilities.The main contents of this paper are as follows:1.An infrared and visible image fusion method based on M-channel oversampled filter banks is proposed.Specifically,the source images are treated as a signal on a regular graph and decomposed into low frequency and high frequency subbands using M-channel oversampled graph filter banks,which represent different frequency component in the graph wavelet domain.Then,using the standard deviation based local properties and the bilateral filter to construct the fusion weight values of the low frequency subbands and the high frequency subbands.The coefficients of each subband are fused according to this fusion rule.Finally,the fused image is obtained by applying an inverse transform to the fusion subband coefficients.The experimental results have shown the effectiveness of the method,both in terms of visual effects and objective evaluation indicators.2.Near infrared,far infrared and visible image fusion based on two-channel oversampled graph filter banks is proposed.The fusion of near infrared,far infrared and visible images combine the target information of near infrared and far infrared images and the details of visible image.It can provide information images for human visual perception and subsequent processing.In order to obtain more detailed signal decomposition,oversampled two-channel graph filter banks on the bipartite graph is used.Then,the fusion rules of low frequency subband and high frequency subband are independently constructed by automatically selecting weighted average or maximum absolute value.Finally,the fusion image is obtained by applying inverse transform to the fusion subbands.The experimental results have shown the effectiveness of the method,both in terms of visual effects and objective evaluation indicators.