一种基于小波和Hough变换的仿射不变性商标检索

A Retrieval of Affine Invariant Trademark Based on Wavelet and Hough Transform

提出了一种基于小波和Hough变换的仿射不变性商标检索方法.根据商标图像色彩单调、纹理简单的特点,首先对商标图像进行二值化处理并做Hough变换;然后,对经Hough变换的商标图像信息进行规范化处理,使其对应的小波变换具有仿射不变性,并用Mexican Hat小波对规范化后Hough变换信息进行奇异点分析.最后根据奇异点的信息进行商标检索.实验结果表明,该方法在检索效率和检索准确率上均高于现在应用较多的仿射不变矩算法.

This paper presents an approach of affine invariant trademark retrieval based on wavelet and Hough transform. First, considering the simpleness of color and texture of trademark, it use binarization and Hough transform to process the original images. Then, affine-invariant property of wavelet transform is achieved by normalizing the information from Hough transform. Next, it use Mexican Hat wavelet to detect the odd points and use the information of these odd points to make trademark retrieval. Finally,experimental results show that this algorithm is more precise and effective than affine moment invariants algorithm.