Image fusion is an important method for improving feature extraction, classification and objective recognition capacity of remote sensing images. In order to compare the effect of different fusion methods, this paper uses the algorithm based on HIS transform and low pass filter, PCA, Highpass Filtering and GramSchmidt to implement fusion experiment on SPOT5 data and TM data to integrate the images in Beijing, and uses four methods such as mean value, standard deviation, information entropy and relative coefficient to evaluate the spatial information and the spectral fidelity. The results show that in the four methods, GramSchmidt fusion method has the most advantage in reserving information and that its spectral fidelity is flat as PCA method. Highpass Filtering fusion method is all good in the retention of information and spectral fidelity but its brightness of the fusion result is worst among the above four methods.