摘要
为了提高图像Delaunay三角形化(Delannay Triangulation,DT)的速度及恢复图像的质量,在文献[1]方法的基础上,对自适应图像Delaunay三角形化的方法进行了结构性的改进.改进后的新方法采用了以三角形、边、顶点为基础的类结构,并以三角形的边描述三角形间的相邻关系.由于每条边本身就是两个相邻三角形的交界,加之类特别适合描述相互关系,使得算法得到较大的简化,运算复杂度也随之减小了.在算法的关键步骤采用双精度计算,提高了网格恢复图像的描述精度.实验结果表明,新方法产生的网格随图像内容自适应变化,在网格生成速度上比文献[1]提高了约1/3,在恢复图像的PSNR(峰值信噪比)上比文献[1]提高约(0.02～0.08)dB.
To speed up the Delaunay triangulation （DT） of the image and improve the quahty ot its reconstructed image, new structural improvements based on our proposed adaptive DT scheme are presented. Based on fundamental data structure of triangular, edge and vertex classes, this new scheme uses the edges of a triangle to record adjacent relations with other triangles. Because each edge is the intersection of two adjacent triangles, as well as the OOP class is good in describing relations of objects, the algorithm is simplified and the computational complexity is reduced. Double precision calculation helps to improve the refinement of the reconstructed image from Delaunay grid. Experimental results show that new scheme can produce grids related to image content adaptively and improves the scheme of reference [1] in grid generating speed about 1/3 and PSNR of reconstructed image about （0.02-0.08） dB.
出处
《中北大学学报（自然科学版）》
EI
CAS
2007年第4期360-365,共6页
Journal of North University of China(Natural Science Edition)
基金
国家自然科学基金资助项目(60472083)
