There are many types of features in heterogeneous images, and the differences between these features are significant; thus, ensuring that different features can be fused into a single image is difficult. To address this problem, a hybrid fusion algorithm for multi-feature fusion between heterogeneous images is proposed in this paper. First, the contour features of the heterogeneous images are extracted by using the local pixel meaning value contrast, and the fusion weight of the contour feature is obtained using a guider filter with the aim of fusing the contour features between images. Second, the edge features of the heterogeneous images are extracted by using the local standard deviation, and then fused by using the local standard deviation similarity between images as the threshold, yielding the edge features of the fused image. Third, the contour-feature fusion and edge-feature fusion images are combined to obtain the low-frequency fusion image. Fourth, the source images are transformed by using two-dimensional variation mode decomposition (2D-VMD), and a detailed-feature fusion image is obtained by using the skewness and an impulse neural network (PCNN). Finally, the spatial domain fusion image is injected into 2D-VMD as the low-frequency fusion image of 2D-VMD, and the fusion result is obtained via inverse transformation of 2D-VMD. The experimental results demonstrate that the proposed fusion algorithm can clearly improve the fusion performance of heterogeneous images, and generate a more informative fusion image.