一种基于Zernike不变矩和改进的Hough变换的图像检索方法

IMAGE RETRIEVAL USING COMBINED DESCRIPTORS OF ZERNIKE INVARIANT MOMENT AND IMPROVED HOUGH TRANSFORM

针对基于内容的图像检索为了解决不变矩只关心对象的区域特征,而忽视了图像边界的缺点,提出Zernike不变矩与改进的Hough变换相结合的方法。改进的Hough变换方法充分利用Hough变换空间的统计信息,对形状的仿射变换具有一定的鲁棒性、易并行处理和良好的几何解析性等优点。该算法有效地融合了图像形状的边缘和区域特征。实验结果表明,在图像检索的综合性能上优于Zernike方法以及基于改进的Hough变换的方法,检索速度快,准确率高。

In light of the drawback of content-based image retrieval that for solving the invariant moment it only pays attention to regional features of the object but overlooks image boundary,we present the method which combines Zernike invariant moment with the improved Hough transform. The improved Hough transform method puts the statistic information of Hough transform space into full play,and has the advantages of having certain robustness,easy for parallel processing and good geometric analyticity on shape＇s affine transform. The algorithm presented in the paper effectively integrates the boundary and regional features of image shape. Experimental results indicate that the proposed method is superior to the Zernike method and the improvement-based Hough transform method in terms of comprehensive performance of image retrieval with high retrieving speed and accuracy.