【作者】 李红；

【导师】 唐国维；

【作者基本信息】 大庆石油学院 ， 计算机应用技术， 2005， 硕士

【摘要】 基于对象的视频编码是以图像处理、模式识别及人工智能为基础,充分利用人类视觉系统的生理和心理特征以及信源的各种性质去掉图像内容的冗余信息,采用基于对象的方法,实现对视频的压缩编码。 本文研究基于对象的视频编码的关键方法和算法,并根据 MPEG-4 视频编码标准的要求在改进一些关键算法的基础上实现了适用于可视电话、视频会议和人机交互等业务中的头肩序列图像的编码方法。 基于对象的视频编码主要包括视频对象的分割和视频对象的编码两部分。本文首先按照基于对象的视频编码的要求并针对头肩图像的特点,对输入图像进行对象分割。然后根据人类视觉系统的生理和心理特性将分割后的视频对象按照不同的重要程度采用不同的编码方法进行编码。通过系统地学习和分析现有的形状编码、纹理编码和运动编码技术,提出了基于二值 a 宏块和运动补偿的形状编码方法、改进的形状自适应离散余弦变换的纹理编码方法和利用块特性的自适应十字搜索模式的运动估计和补偿算法。实验结果表明,本文的改进算法比较有效,避免了大量的冗余计算,在保证图像质量的同时提高了编码效率,在基于对象的视频编码技术研究中取得了一定的理论和实用价值。更多还原

【Abstract】 Object-based video coding is based on image processing, pattern recognition and artificial intelligence. The physiological and psychological characteristics of human visual system and the properties of signal are used to get rid of redundant signals of the image. Object-based method is adopted to implement video compress coding. In this paper, the key method and algorithm of object-based video coding is researched. The head-shoulders video sequence coding is implemented which in many applications such as videophone, videoconference and human-computer interface, etc. This method is based on the need of MPEG-4 video coding standard and the improvement of key algorithm. Object-based video coding mostly includes video object segmentation and video object coding. First, the input images are segmented according to the demand of object-based video coding and the characteristics of the head-shoulders video sequences. Second, the different video object is coded by different means based on human visual system. Some improved methods are put forward through systematical studying and analyzing of current shape coding, texture coding and motion coding techniques. These new methods include BAB and motion compensation based shape coding, improved SA-DCT texture coding and block based adaptive rood pattern search motion estimation and compensation method. The result of experiments indicates that these improved methods are relatively efficient. These methods increase the coding efficiency and guarantee the quality of images by avoiding a great deal of redundant calculation. This research obtains certain academic and practical value of object-based video coding techniques.更多还原