Abstract: The feature description and evaluation is the critical phase in image mining. While briefly introducing the concept of the Zernike moment, this paper points out that the shape feature of the image by the Zernike moments set can be described. After talking about the image reconstruction technique based on Zernike moment, accuracy of the Zernike moments shape feature set via the dissimilarity degree and image reconstruction rate between the original image and the reconstructed image can be evaluated. The experience results demonstrate the feasibility.

Key words: Image mining, Zernike moments, Shape feature set, Image reconstruction

摘要： 在图像挖掘中，最关键的步骤是提取图像特征并对之进行描述和评价；在介绍Zernike矩的基础上，指出可以使用Zernike矩集描述图像的形状特征；根据Zernike矩逆变换，可以得到基于Zernike矩形状特征集的图像重构技术，从而通过重构图像与原图像的相异度和重构率来对Zernike矩特征集描述图像形状特征的精确度进行评价；实验结果证明了基于Zernike矩描述图像形状特征与基于图像重构进行评价的可行性。

关键词: 图像挖掘, Zernike矩, 形状特征集, 图像重构