Abstract: In many application scenarios, a low complexity video encoder is needed. the combine of distributed video coding and compressive sensing technique is fit for these scenes. So a new video coding scheme, distributed video compression coding, appears. In some distributed video compression coding schemes, the video frames are encoding independently at the coder and decoding jointly at the decoder. Specifically, the key frames decode independently and non-key frames decode with the help of the side information which is generated by key frames. This ignores the correlations between non-key frames. This paper proposes a new distributed video coding scheme, dividing the non-key frames into the main non-key frames and sub-non-key frames. The main non-key frames decode with the help of the side information which is generated by key frames. The sub-non-key frames make use of the adjacent main non-key frames to predict the measurement firstly. Then The sub-non-key frames decode with the side information which is generated by key frames. The experimental results show that the proposed scheme improves the PSNR of non-key frames there 2dB ~ 4dB .