In　this　paper,　a　fast　face　detection　algorithm　for　JPEG2000　color　images　is　presented,　which　combines　both　color　and　texture　information　in　order　to　find　a　good　balance　between　speed　and　detection　reliability.　The　algorithm　is　designed　to　work　directly　on　the　wavelet　compressed　domain　which　possesses　the　following　characters:　First　of　all,　the　multi-level　gradient　energy　presentation　of　face　pattern　is　proposed,　which　not　only　can　highlight　the　facial　parts　in　possible　face　patterns,　but　also　can　address　effectively　the　problem　of　unknown　size　in　face　detection　in　compressed　domain　and　therefore　avoid　the　complex　resolution　transform　in　arbitrary　ratios　;　secondly,　the　skin-color　model　in　YCbCr　space　is　ameliorated　to　improve　the　reliability　of　skin　segmentation;　finally,　a　hierarchical　detector　which　integrates　the　simply　rule-based　classifiers　and　complex　neural　network　based　classifier　is　designed　to　further　improve　the　processing　speed.　Experimental　results　show　that　the　proposed　scheme　is　efficient　and　effective.