一种基于运动对象的形状编码新算法

本文提出一种基于运动对象的形状编码新方法,称之为基于曲率尺度空间CSS(Curvature Scale Space)的自适应算术形状编码算法(CSSAS).本算法主要分为两种编码模式:帧内模式和帧间模式.在帧内模式中,我们在对CSS算法进行改进的基础上,对任意形状对象的形状信息进行特征点的分层提取;并用自适应算术编码算法对提取的特征点进行编码.在帧间模式中,提出了一种基于曲率尺度空间图(CSSI)的任意形状对象的运动估计算法.上述运动估计/补偿后所获得的形状曲线的匹配部分使用基于弧长索引的压缩编码算法,而对于不匹配部分,采用与帧内CSSAS算法相同的方法进行压缩编码.实验结果表明,本文提出的CSSAS算法与MPEG-4校验模型中基于上下文的算术形状编码算法(CAE)相比,在帧内模式时,CSSAS比CAE压缩比提高约25%,在帧间模式Dn较大时,CSSAS比CAE压缩比明显提高,而在重构形状的主观质量上,CSSAS这两种模式均优于CAE.

A New Motion-Object-Based Shape Coding Algorithm

In the paper,a new motion-object-based shape coding called CSSAS (curvature scale space adaptive arithmetic shape coding) is proposed.There are two coding modes,the intra and inter modes.In the intra mode,the modified CSS theory is used to extract feature points of the shape curve of arbitrary shape video object(VO) at different scales.All of the feature points are encoded by adaptive arithmetic encoding.In the inter mode,a CSSI-based (Curvature scales space image-based) motion compensation/estimation algorithm of arbitrary shape VO is proposed.The matched segments of shape are coded by using arc length index algorithm and the mismatched segments are coded by using intra CSSAS algorithm.Simulation results show that the proposed method is much better than CAE of MPEG-4 VM in coding efficiency (in inter mode,Dn≥threshold) and subjective quality for both intra and inter modes.