Abstract: A gesture extraction algorithm with low complexity and high accuracy is presented in this paper.This algorithm meets the needs of real-time human-computer interaction,solves the problem of gesture extraction which has nothing to do with skin color in different light conditions and different backgrounds.And it is suitable for hardware implementation.The algorithm extracts the gesture based on the row-by-row basis in binocular images.Firstly,edges in both images are detected and images are segmented by the edges so that the gesture contour can be obtained.Secondly,the gesture is extracted by a matching algorithm.Finally,background noise are eliminated and accurate gesture comes out.In this paper,the binocular images are captured by two same web cameras in the indoor lighting conditions.The experimental results show that the algorithm is able to extract various gestures quickly and accurately in complex background,and is more robust compared to traditional gesture extraction algorithm based on skin color.