Human face detection in complex background with conventional methods usually gets unsatisfied results.In this paper，a new method based on HS-CbCrCg color space and Adaboost algorithm is presented.First，light compensation and image enhancement are implemented for the input images under different illumination conditions，then color discrimination is carried out by using the established skin color model in HS-CbCrCg color space.Secondly，human face information in the skin region is obtained by using the improved Adaboost algorithm.Finally，the confident face can be discriminated by repeat tests on the key regions.The experimental results show that compared to the traditional face detection methods，the proposed method achieves excellent performance，it is fast to process and has lower missed detection rate and lower false detection rate.