This　paper　illustrates　face　recognition　based　on　spanning-tree　kernels　to　improve　the　accuracy　of　face　recognition,　defines　the　concept　of　max　spanning　tree,　seven　spanning-tree　kernels　and　three　complicated　kernels　based　on　spanning-tree　kernel　and　histogram　intersection　kernel.　Experimental　result　shows　that　spanning-tree　kernels　have　significantly　higher　classification　accuracy　than　walk-based　kernels.