Visible and Infrared Image Fusion Based on Convolutional Sparse Coding with Gradient Regularization

The purpose of visible and thermal infrared scene fusion is to generate a synthetic image, in which clear thermal target and pleasant visual background can be obtained simultaneously. Standard convolutional sparse coding is an effective method to solve the problem of detail conserve and sensitivity to registration errors in sparse domain fusion method. However,partial infrared-visible fusion results based on standard convolution sparse coding have lower contrast as different imaging modalities of infrared-visible images.The gradient regularization of convolutional sparse coefficient graph is introduced into convolutional sparse coding and a new visible-infrared image fusion method is proposed. Experimental results demonstrate that our method can achieve clearly fusion performance in terms of both objective and visual.