一种新的基于链码特征的图像检索算法

Novel Algorithm for Image Retrieval Based on Chain Code

在分析采用Freeman链码进行形状描述的基础上,提出了两种链码空间分布特征的提取算法:链码分布矢量及链码相关矢量,同时对这两种方法的尺度、旋转、平移不变性进行了分析和验证;针对两种算法,分别设计了有效的相似性度量策略;最后结合链码直方图进行图像检索。由于该方法在进行图像检索时既考虑了链码的统计特征又包含了其空间分布特性,因此取得了比传统方法更好的检索效果,试验结果也证明了该算法的有效性。

Based on the analysis of chain code in shape representation, two novel shape descriptors, named chain code distribution vector and chain code coherence vector, are introduced to express the spatial feature in the chain code. These two descriptors have the advantages of being invariant to the position, rotation and scaling of the image content and have nothing to do with the start point of the chain code. Combined with chain code histogram, two different matching methods are presented to measure the similarity of shape information. It is clear that both the statistical feature and the spatial feature of the chain code are considered in the new methods. Experiment results show that our methods give better performance than the traditional methods.