One novel multi-band image fusion algorithm based on neighborhood limited empirical mode decomposition(NLEMD)was proposed.Firstly the images were decomposed by NLEMD,the parts of intrinsic mode functions(IMF's) and remnant correspondingly to each image were obtained.Then in these IMF's and remnant images,the pixel which has the maximal energy was selected for the result IMF images and remnant image.At last,the final result image was reconstructed.With the help of NLEMD,the image's detail can be extracted and the final fusion image is clearer than the source images.Experiments show that the new algorithm has more advantages in achieving high frequency details of source images than the fusion algorithm based on wavelet analysis.