基于链码分布特征及相关性的轮廓描述与检索

Contour representation and retrieval based on spatial feature and relativity of chain codes

在分析Freeman链码进行边缘描述的基础上,提出采用链码熵来描述链码的统计特征、采用链码空间分布熵来描述链码的空间分布特征、采用链码相关熵来描述链码间的相关性特征,并结合这三种特征进行图像检索。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征及相关性特征,同时上述三种描述子又具有尺度、旋转、平移不变性及链码起点无关性,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 cedes in shape representation,three novel contour descriptors,called chain code entropy（CCE）, chain code spatial distribution entropy（CCSDE） and chain code relativity entropy（CCRE） are introduced to express the statistical feature, the spatial distribution feature and the relativity feature in the chain cedes. These there descriptors have the advantages of being invariant to translation, rotation and scaling of the contour and have nothing to do with the start point of the chain codes. Based on the three descriptors, a novd image retrieval algorithm is proposed. Experiment resuits show that our method give better performance than the traditional methods.