Abstract: Proposed a dynamic independent component analysis domain fusion scheme based on the redundancy and complementary information of the source image analysis. Based on analyzing the traditional ICA domain grayscale image fusion algorithm, the dynamic ICA method was proposed that choosing the specific image blocks from the source images as training image blocks, which made full use of the redundancy and the complementary information of the source images without considering the type of the image and the ICA domain transform function changes with the source images dynamically. The proposed method was compared with N.M 2007 method and dual tree complex wavelet transform fusion method. The simulation result shows that the proposed method enhances environment applicability, reduces computation, establishes a more comprehensive and accurate expression of the scene information, further improves scene perception.