Visual-hint Boundary to Segment Algorithm for Image Segmentation

摘要

Image segmentation has been a very active research topic in image analysisarea. Currently, most of the image segmentation algorithms are designed basedon the idea that images are partitioned into a set of regions preservinghomogeneous intra-regions and inhomogeneous inter-regions. However, humanvisual intuition does not always follow this pattern. A new image segmentationmethod named Visual-Hint Boundary to Segment (VHBS) is introduced, which ismore consistent with human perceptions. VHBS abides by two visual hint rulesbased on human perceptions: (i) the global scale boundaries tend to be the realboundaries of the objects; (ii) two adjacent regions with quite differentcolors or textures tend to result in the real boundaries between them. It hasbeen demonstrated by experiments that, compared with traditional imagesegmentation method, VHBS has better performance and also preserves highercomputational efficiency.