The multi-reference frames technique is employed in H.264
which can improve prediction precision and meliorate video pictures quality greatly.However
the feature calls for higher computational complexity simultaneity.In order to reduce the complexity of H.264 encoder
a fast multi-reference frames selection algorithm is presented in this paper.By considering the correlation and center-biased characteristic of motion vectors in multi-reference frames adequately
the proposed algorithm makes selection for multiple references in advance
thereby it can effectively decrease the numbers of reference frames making full search and improve the coding efficiency.Experimental results show that the proposed methods can improve the speed of motion estimation module by 32.14% on average with less performance degradation and bit-rate increase.