Image Fusion Based on Masked Online Convolutional Dictionary Learning with Surrogate Function Approach

To avoid loss of detail information and low contrast of fusion results caused by multi-scale transform image fusion algorithm, an image fusion based on masked online convolutional dictionary learning with a surrogate function approach is proposed by introducing image fusion into online convolutional dictionary learning algorithm. The dictionary learning algorithm is used to obtain the over-complete dictionary filter, and then convolutional basis pursuit denoising algorithm is used to obtain the high-frequency and low-frequency sparse coefficients. The fusion image is finally reconstructed. To prove the superiority of our proposed algorithm, six groups of representative infrared and visible images are applied to our method and three comparative methods. The experimental results show that the fusion image of the proposed algorithm achieves good results in subjective and objective quality evaluation. Compared with the JSR-based method, NMI, QTE, and QNCIE increased by 28.04, 19.41, and 0.14% averagely.