A Novel Fusion Strategy and Convolutional Sparse Coding for Robot Multisource Image Fusion

Image fusion refers to the fusion of images collected by multiple sensors about the same target or scene into one image by image processing technology. In this way, the advantages of multiple sensors can be effectively utilized to obtain more comprehensive feature information of the target or scene, which is conducive to human eye observation and subsequent recognition and processing. The traditional fusion methods are easy to lose the image details, resulting in poor fusion effect. Therefore, this paper proposes a multisource image fusion method based on convolutional sparse coding and a novel fusion strategy. Firstly, the image is decomposed into low-rank and sparse through low-rank decomposition. Then, the sparse part is convoluted and decomposed to obtain a set of sparse filter dictionaries, which are applied to image fusion by convolutional sparse coding. The regional energy-Cauchy fuzzy function rule is adopted for low-rank components. The regional Laplace energy is used for sparse component. Finally, the weighted average method is used to obtain the final fusion result. Experimental results show that the proposed method achieves good results in terms of visual effects and objective evaluation indexes.