An evaluation model for different-source image fusion quality based on FNN is proposed.Subjective evaluation conclusion sample sets of fusion images are considered as output of fuzzy expectation.Several classical objective evaluation indexes are fuzzed by Gaussian membership function as network input samples.Related parameters of evaluation index weight and membership function are generated by network learning.Momentum factor is adopted to improve network learning efficiency.Experimental results show that the evaluation results are reasonable to human eyes.The uniformity ratio of subjective and objective evaluation can reach a high rate,which provides a valuable method for the realization of automatic fusion image evaluation.