after analysis and comparison of different matching algorithms, a kind of real-time image splicing algorithm is proposed basic on the combination of gray level and characteristic points. Firstly, the algorithm performs self-adaptive adjustment to gain required size, secondly, determines matching direction and general position with characteristic marginal points and then optimal matching points with related method of gray, thirdly, carries out blending treatment for overlaying images with weighted mean. Result of experiment indicates that algorithm can make precise splicing on real-time and effective basis.