To　solve　the　problems　of　sensitivity　to　little　color　change,　huge　computation,　low　effect　that　exist　in　content-based　image　retrieval　methods,　a　new　method　is　proposed.　This　retrieval　method　is　based　on　image　color　and　texture　features.　First,　each　part　of　HSV　color　space　is　quantized　using　non-uniform　quantization.　Then,　to　eliminate　the　quantization　error,　adding　each　two　neighbor　values　in　the　histogram　to　form　a　three-dimensional　color　histogram.　A　corresponding　histogram　intersection　matching　method　is　proposed　at　the　same　time.　Moreover,　a　texture　parameter　is　constructed　to　remove　irrelevant　images.　Experiments　show　that　the　new　method　is　effective　to　color　image　retrieval.　It　is　also　low　in　computation　and　robust　to　slight　color　changes.　©　2012　Springer-Verlag　GmbH.