基于特征的图像网格生成方法

Approach to feature-based image mesh generation

基于网格形变的图像缩放算法是目前的一个研究热点。适当的图像网格表示是这类算法成功的关键之一。提出一种基于图像特征的三角形网格生成算法。提取图像分割形成区域的边缘特征点,与图像四条边界上均匀分布的点一起,作为改进的Dart-throwing算法的初始点集。用距离变换计算每个像素到最近边缘线的距离,作为Dart-throwing算法的控制参量;所生成的网格点集接近边缘线时密集,远离边缘线时稀疏。最后用Delaunay算法形成图像的三角形网格表示。实验结果表明,所生成的三角网格较好地体现了图像的结构特征,并且有效减少了网格点数目,有利于提高后续算法的处理效率。

Image resizing based on mesh deformation is becoming ever more important with the proliferation of display units with different aspect ratios. Appropriate mesh representation for the input image is a key aspect to the success of this kind of algorithms. An effective approach to feature-based image triangular mesh generation was proposed. The feature points extracted from the image edges and the uniformly located boundary points were used as the initial points set of the modified dart-throwing algorithm. The distance, each pixel to the closest edges, was computed by the distant-transformation of the edge image. It was used as the control parameter of dart-throwing algorithm. The Delaunay algorithm was then exploited to create a triangular mesh. The experimental results illustrate that the generated meshes well describe the edge-structure of the input images. Moreover, the post-processing algorithms will be more efficient due to the smaller number of the mesh nodes.