Abstract: Video object segmentation (VOS) is easily affected by the object fast motion, extrinsic occlusion, etc., so the high-precision VOS remains challenging in the area. Optical flow can be used to improve object segmentation, but the precise estimation of optical flow around the moving boundary is hard to be fulfilled, which affects the unveiling of VOS performance based on optical flow estimation. To overcome the above limitations, this paper combined the motion contour information extracted by the retinal macro-cell pathway model to assist in calculating the optical flow of the motion boundary region, and combined with the traditional VOS method foreground-background segmentation to alternately update the optical flow and segmentation. Experimental results on several public benchmarks, i.e.KG-*3, DAVIS-2016, SgeTrack-v2 and YouTube-Objects, show that the proposed method improves the average segmentation accuracy by 2.2, 1.3 and 1.9 percentage points respectively, compared with the baseline method.