A scene change detection algorithm based on non-connected points is proposed to solve the problems of scene changes existed in videos

which can improve the coding performance. The complexity of the proposed algorithm is low and scene change detection is processing at the same time of the motion estimation. Scene changes can cause different GOP(group of pictures)sizes

which can result in GOP with a short size. We propose an improved adaptive GOP motion compensation temporal filtering

which can effectively avoid the short GOP problem resulting in a decline of coding perfor- mance. For each GOP generated by scene change detection

a frame-layer rate control algorithm based on a rate distortion model is proposed

using the relation between the distortion and the rate and the complexity of the video frame to optimize the rate distribution

aiming at improving the quality of the reconstructed video. The experimental results show that the proposed algorithm can achieve better performance than other rate control algorithms.