Content-based　image　retrieval　represents　images　as　N-dimensional　feature　vectors.　Similar　image　retrieval　is　computed　over　these　high　dimensional　feature　vectors.　A　sequential　scan　of　the　feature　vectors　for　a　query　method　is　costly　for　a　large　number　of　images　when　N　is　high.　The　search　time　and　search　space　can　be　reduced　through　indexing　the　data.　In　this　paper　we　proposed　a　hierarchical　clustering　tree　for　building　indices　of　the　image　database.　Experiment　shows　that　the　hierarchical　tree　has　the　attribute　of　quick　response　time　when　retrieve　image　based　on　the　query　and　also　provide　a　convenient　structure　for　users　to　browse　the　database.