This paper presents a new metric to assess the performance of the image fusion algorithm. The proposed metric is based on image phase congruency, which provides an absolute measurement of image feature with a numeric value ranging from zero to one. By comparing the local crosscorrelation of corresponding phase congruence maps of input images and fusion output, the quality of the fused result is assessable without a perfect reference. The preliminary experimental results on multi-focused image pairs demonstrate the efficiency of this approach.