In　this　paper,　we　introduce　a　new　method　for　digital　camera　identification　from　its　color　images　using　image　sensor　noise.　We　first　compute　the　two　noise　reference　patterns　by　averaging　the　noise　component　from　two　groups　of　color　images　taken　with　a　camera.　Then　we　use　canonical　correlation　analysis(CCA)　to　calculate　the　projection　directions　of　the　two　noise　reference　patterns.　Finally,　we　calculate　the　correlation　coefficient　between　the　projection　of　the　noise　from　a　specific　color　image　onto　one　projection　direction　and　the　projection　of　one　of　noise　reference　patterns　onto　another　projection　direction,　then　use　this　coefficient　to　decide　whether　the　specific　color　image　was　taken　by　the　camera　or　not.　Experimental　results　show　that　the　presented　method　provides　higher　accuracy　than　other　methods　on　the　condition　of　using　a　few　images　to　compute　reference　pattern.　©　2008　IEEE.