摘要： 为验证贝塞尔-傅里叶矩在图像检索中的检索性能及图像识别能力，在ICU和ALOI图像库中，提取贝塞尔-傅里叶矩特征，进行检索和分类实验。实验结果表明：贝塞尔-傅里叶矩对纹理图像不仅具有较高的检索率，而且RPP曲线能够达到理想状态；对彩色图像形状检索具有较高的查准率和查全率，对图像存在的光照变化和几何变化尺度旋转具有较强的鲁棒性。相同阶数的贝塞尔-傅里叶矩特征，在KNN分类实验中可以获得更高的分类准确率。

关键词: 贝塞尔-傅里叶矩, 图像检索, 图像分类, 旋转不变性, 图像重建

Abstract: In order to test the retrieval performance of the Bessel-Fourier Moment(BFM), the feature descriptor extracted with the proposed algorithm is used for image retrieval and classification in different databases: ICU and ILL databases. Experimental results show that the higher retrieval precision rate can be achieved, and the curve of RPP amounts to an ideal state for invariant texture retrieval. The performance of BFM is better for shape recognition, and has strong robustness to varying illumination and rotational invariance in color images retrieval. The BFM achieves higher correct classification rate than conventional descriptors with the same order.

Key words: Bessel-Fourier Moment(BFM), image retrieval, image classification, rotation invariance, image reconstruction

中图分类号: 

• TP3