Multisource image fusion technology is widely applied in a variety of fields such as remote sensing, computer vision, medical diagnosis and military surveillance. In most instances, multi-spectral and panchromatic images can provide more complementary information for feature extraction. Two kinds of images are used for target recognition. Algorithms based on gradient for image fusion only consider high-frequency information changes of the images, and neglect the richness of high-frequency information. To solve this problem, a new kind of self-adaptive rule and algorithm based on gradient and energy is proposed in this paper by analyzing the internal relations between the statistics and the content of the images. The rule and algorithm take into account the changes in high-frequency information and the richness of high-frequency information, and guarantees the richness of image information. Experiments show that the fusion rule and algorithm can significantly preserve all useful information from primitive images and effectively recognize targets with higher recognition accuracy.