The　obstacle　recognition　and　segmentation　in　image　sequences　has　become　one　of　the　key　technologies　of　the　robot　obstacle　avoidance.　In　this　paper,　a　new　method　to　obstacle　regions　extraction　from　images　for　mobile　robots　is　proposed.　In　the　proposed　method　a　Pulse　Coupled　Neural　Network　(PCNN)　and　an　improved　Chan–Vese　(C-V)　level　set　algorithm　are　applied　for　obstacle　recognition　and　classification　through　a　robotic　vision　system.　Then　A*　search　algorithm　is　used　to　achieve　path　planning　and　graph　traversals.　The　result　shows　that　the　method　can　efficiently　extract　obstacles　region　in　the　field　of　view.　Furthermore,　the　validity　and　practicability　of　the　proposed　approach　was　validated　by　a　lot　of　experiments　on　the　mobile　robot　Pioneer3-DX.　©　2015　Taylor　&　Francis　Group,　London.