The covariance intersection algorithm is a distributed fusion estimation method obtained by optimizing certain objective functions, which provides a new idea for image fusion enhancement.An image fusion method based on the Covariance Intersection(CI) algorithm and the local gradient extrema based Bidimensional Empirical Mode Decomposition(BEMD) is proposed.To overcome the inadequacy of traditional BEMD in obtaining image details and to include more detailed structure features in the image, according to the strong ability of the gradient to mine the detailed information of the image, local gradient extrema are selected by using four two-dimensional extremum conditions.Then, empirical mode decomposition of the image is carried out and the IMF is determined.Then, the one-dimensional covariance intersection algorithm is extended to 2D signals and image fusion.The optimal linear weighting matrix is computed by minimizing the F-norm of the 2D covariance intersection matrix of each IMF.The enhanced fusion image is obtained by using inverse reconstruction.The simulation results show that, compared with the traditional image fusion algorithm, the proposed method has stronger detail capture ability and better clarity.