Abstract: This paper presents a novel approach for eye detection based on an improved Adaboost algorithm in combination with the SVM （support vector machine） technique.A cascade classifier based on an improved Adaboost algorithm,which uses much fewer features than the conventional Adaboost algorithm,is used for efficient face and eye detection with complex background,followed by a non-linear SVM classifier to furthe r refine the eye location in consideration of the rich geometric information of a face.Experimental results show that our proposed method is capable of achieving both good accuracy and real-time performance.