Abstract: A novel fine granularity scalability coding framework with hybrid open-close loop structures based on key reference pictures was proposed.In the framework,the open-loop structure with one prediction loop is adopted to code non-key frames to achieve the best coding efficiency;some key frames exploiting the close-loop structure with two prediction loops are periodically inserted to control drift caused by the open-loop structure.To optimally exploit the bandwidth,an optimized bit-stream extraction method was proposed to give more priority to the enhancement layer of key frames than that of non-key frames.Simulation results show that the proposed framework outperforms AR-FGS(FGS with adaptive reference) in JVT SVC(joint video team scalable video coding) by about 1 to 2 dB in a very wide rang of bit rate.