Aiming at the performance evaluation of image fusion, this paper discusses the existing criteria and the guidelines for choosing performance parameters, and then a new method is proposed to index the fusion algorithms on their fused images. In the method, each of the fused images is used to extract a comparative sequence. Meanwhile, a reference sequence is constructed by combining all the fused images involved. Then an overall performance index is obtained from the gray relational degrees between the two kinds of sequences. The interest of the method lies in the fact that it does not require a ground-truth or reference image in forming the reference sequence and is inexpensive in processing time and storage memory. Experimental results show that the method is effective, and its potential application includes intellichoice and intellectual control.