关键词: 多视点视频编码, 多参考帧快速选择算法, 双目恰可察觉失真（BJND）

Abstract: To reduce the extremely high encoding complexity in multiple reference frame prediction, a fast algorithm for multiple reference frame selection based on Binocular Just Noticeable Difference（BJND） is proposed. By analyzing the statistical features among multiple reference frame selection and BJND, the necessity for full searching among all reference frames is judged. When BJND value is larger than the thresholds, the first reference frame in the temporal or inter-view prediction of the lists is selected as the best reference frame. Experimental results show that the proposed algorithm can reduce averagely 76% of the encoding time compared with JMVC while the rate distortion performance nearly remains the same.

Key words: multiview video encoding, fast algorithm for Multiple Reference Frame（MRF）selection, Binocular Just Noticeable Difference（BJND）