The morphological characteristics of retinal vessels play a significant role in the diagnosis and screening of fundus-related diseases. However, due to the diverse scales and shapes of retinal vessels, as well as large differences in contrast, the detection of fine blood vessels becomes a challenge. To address this problem, retinal vessel segmentation based on image fusion of B-COSFIRE (Bar Selective Combination of Shifted Filter Responses) and phase congruency is proposed. The feature response map was obtained using wavelet transform based image fusion of the B-COSFIRE filtering response and phase congruency of the preprocessed green channel. And then retinal vessel segmentation was implemented by the fuzzy C-means clustering algorithm. The experimental results show that the average accuracy, sensitivity and specificity of the proposed method on DRIVE database is 94.50%, 69.63%, 98.16% respectively. The proposed method exhibits good segmentation performance, relatively complete segmented vascular network, and improving the sensitivity to detection of small blood vessels.