Albeit great success has been achieved in image fusion research, Fusion Image Quality Assessment (FIQA) is a still puzzling problem to be solved. In this paper, Deep learning (DL) is explored for FIQA task and a new Multi-Focus Image Fusion (MFIF) assessment method is proposed. We innovatively formulate the multi-focus FIQA task as a classification-learning based focus similarity comparison process, including focus property classification, focus similarity comparison and final assessment and analysis. Extensive objective contrasts and subjective comparisons demonstrate the correctness and specificity of our method.