摘要： 在动画形象的形状检索基础上，通过轮廓分解及Hausdorff距离的改进，提出一种动画形象素材的形状检索方法。该方法对形象进行前期预处理，得到形象二值闭合边界点集；根据边界围成区域的主惯量轴，将形象图片归一化；利用角点检测到的角点对图像的边界分割，每段弧线等距取相同的采样点数，从而将得到的采样点作为动画形象描述特征点集，利用加权Hausdorff距离测量库中图像与示例图像的距离。实验结果表明，与目前已有方法相比，该方法具有较高的检索精度。

关键词: 轮廓分解, 主惯量轴, Hausdorff距离, 形状检索, 角点检测

Abstract: This paper proposes a shape retrieval method of animation image material by decomposing the contour and promote the Hausdorff distance. The materials are pre-processed in order to get the point aggregate of closed boundary, and the images are normalized by using the axis of least inertia of the area of the closed boudary. It decomposes the contour by corner detector, gets the same number of the points from each arc in order to descript the characteristic of the cartoon characters, measureing the distance of the sample by weighted housdorff distance. Experimental results show this method has higher retrieval precision, compared with existent methods.

Key words: contour decomposition, axis of main inertia, Hausdorff distance, shape retrieval, angle point detection

中图分类号: 

• TP391