边缘模型基背景区的Delaunay三角剖分压缩方法

Delaunay Triangulation Compression of Image Ambient Region in Edgemodel Based Method

在边缘模型基活动图象压缩编码方法中，图象初始帧被分为激励区和背景区．在每个图象组中，初始帧背景区的静态压缩是整个方案的重要基础．本文提出了一种采用Ｄｅｌａｕｎａｙ三角剖分（ＤＴ）的表面描述算法，首先根据伪逆矩阵理论，采用最小二乘空间预测方法填充初始帧中的激励区部分，然后用ＤＴ算法剖分整个初始帧，最后获得少量待传特征点．传输时保证特征点收发的同序，克服了四点以上共圆带来的退化现象．实验表明，在压缩率近１６倍时，重建图象质量仍可保持在３０ｄＢ以上，主观图象质量也较良好．

Initial image frame is divided into sensitized region and ambient region in edgemodel based method of moving image compression coding. In each image group, the still compression of ambient region is the vital foundation of the entire scheme. In the paper, we presents a surface description using Delaunay Triangulation (DT). First, accroding to the pseudoinverse theory, we use space prediction by least square to fill the sensitize region of the initial frame. Then, use DT to triangulate it. Thus, we obtain only few feature points for transmission. To avoid the reconstuction distortion caused by degeneracy, we arrange the bit stream of feature points by forming order. Experiment shows that the reconstructed image can keep at up 30 dB when the compression ratio is 16, simultaneously the subjective quality is also better.