摘要： 传统的小波模极大值结合不变矩法是一种有效的基于内容的商标检索方法，但其进行不变矩计算的边缘定位不够精确，抗噪性不够强，借鉴信号奇异性检测中模极大值线搜索算法思想予以改进，并用VC++编程实现。检索结果表明，该改进算法在保持良好平移不变性、镜像不变性和扭曲不变性的基础上有效提高了查全率和查准率。

关键词: 商标图像检索, 小波模极大值, 不变矩

Abstract: The traditional algorithm which combined wavelet module maxima and invariant moment is an effective algorithm for content based trademark image retrieval. Aiming at the inaccurate orientation and weak antinoise ability in the edge for the invariant moment calculation, This paper uses the method to track the wavelet transform amplitude maxima to mend it. Searching by the images, it found the modified algorithm remarkably improve the recall ratio and the precision ratio when the image have moved, circumrotated, twisted.

Key words: Trademark image retrieval, Wavelet module maxima, Invariant moment