Current Issue Cover


摘 要
目的 针对仿射变换下形状匹配中存在的描述子对形状的描述能力不足,以及描述子计算耗时大的问题,改进基于所有图像点投影的方法,提出一种利用轮廓计算投影面积的仿射形状匹配算法。方法 该算法分为粗匹配和精匹配两个阶段。粗匹配阶段,首先统计轮廓投影面积作为特征点描述子;然后利用动态规划蚁群算法匹配两幅图片公共特征点序列,并将该特征点序列记为对应的新特征点;最后采用该新特征点划分目标曲线。精匹配阶段,采用小波仿射不变描述子,对粗匹配阶段匹配代价最小的5%的目标进行对应曲线匹配,从而达到对仿射目标的识别。结果 算法的平均检索速度比传统基于形状投影分布描述子提高44.3%,在MPEG-7图像库上的检索效果为98.65%,在MPEG-7仿射图像库上的综合评价指标比传统基于形状投影分布描述子高3.1%,比形状上下文高25%。结论 本文提出的算法匹配效果好,效率高,抗噪性强,能有效的应用于仿射形状匹配与检索领域。
Affine shape matching using feature combined with contour and shape

Zeng Jiexian,Gu Ruiyu,Fu Xiang,Len Lu(Institute Of Computer Vision of Nanchang Hangkong University)

Objective In order to improve the computation speed and the discrimination ability of descriptor in the process of shape matching, a affine shape matching method using the projection area which calculate by contour is proposed. Method The algorithm can be divided into coarse and fine match two stages. The coarse matching stage. Firstly, statistics contour projection area as feature point descriptor. And then, ant colony algorithm is employed in match the public feature point sequence in two pictures. Finally, divide the target curve by the public feature point sequence. In precise match stage, Using wavelet transform coefficient construction wavelet local invariant describes the target curve segment to matching the 5% target with minimum cost of the first step, so as to achieve the recognition of affine target. Result The average retrieval rate of this algorithm higher than traditional shape projection distribution descriptor 44.3%. Retrieval result on the MPEG - 7 image library is 98.65%, The comprehensive evaluation index on the MPEG-7 affine image library is higher than traditional shape projection distribution descriptor 3.1%, 25% higher than the shape context. Conclusion The algorithm proposed in this paper has good matching effect, high efficiency and strong noise immunity, and can be applied to affine shape matching and retrieval effectively.