Image　thresholding　segmentation　based　on　entropy　is　classical　method.　Time　cost　of　image　thresholding　segmentation　method　based　on　maximum　entropy　and　enumeration　is　unacceptable,　so　that　the　genetic　algorithms　is　adopted　to　improve　efficiency.　However,　the　performance　of　image　thresholding　segmentation　based　on　traditional　genetic　algorithms　is　not　satisfied　because　of　the　premature　convergence.　Therefore,　we　propose　oriented　genetic　algorithm　to　increase　speed　and　success　rate.　Oriented　genetic　algorithm　includes　an　oriented　crossover　operator　which　directs　generation　of　offspring.　The　blindness　of　genetic　algorithm　is　reduced　and　efficiency　of　optimization　is　improved　due　to　introducing　oriented　crossover　operator.　The　proposed　method　is　compared　with　enumeration　method　and　standard　genetic　algorithm　in　image　segmentation　experiment.　Experimental　results　show　that　performance　of　proposed　method　is　better　than　traditional　methods.