视频对象形状错误隐藏技术研究

Research on Video Object Shape Error Concealment Technology

视频对象(VO)的形状信息对于正确解码至关重要,当VO的形状发生错误时,为了正确解码纹理和运动信息,需要对其进行错误隐藏。主要介绍VO形状错误隐藏的常用技术,并进行对比实验和讨论。结果表明:当相邻VO间运动较小时,时域法比空域法能更好地保持VO的形状;空域法性能稳定,不受VO间的运动影响,当帧间变化较大时,时域法的参考VO失去参考价值,只能选择基于空域的方法。可见,时域法和空域法各有利弊,一个好的解码器应能实现两种错误隐藏,根据具体情况选择或将两者相结合。最后指出形状错误隐藏今后的研究方向和前景展望。

Shape information of Video Object（VO） is most important for correctly decoding. Shape error should be concealed to decode texture data and motion data correctly. Traditional technologies of shape error concealment are summarized in this paper,comparison and discussion are described. When the shape data does not change that much in consecutive time instants,the temporal methods outperform the spatial ones. The spatial methods are the only choice when the previous frame does not contain a shape tat is related to the current shape. Ideally, the decoder should have access to both types of techniques and choose between them or combine them, according to the situation at hand. Finally, the possible research directions of shape error concealment are pointed out.