基于多纹理特征的商标图像检索

提出了一种基于多纹理特征的商标图像检索方法。首先对图像纹理特征进行分析，从人眼视觉角度选用粗糙度和方向性这2个纹理特征量；从统计分析的角度出发，基于图像灰度共生矩阵描述了二阶矩、熵、对比度和均匀性这4个纹理特征量。这6个纹理特征量从不同角度刻画了图像特征，对其归一化后用欧氏距离进行图像相似性度量。通过实验，证明了该方法可以取得比较满意的检索结果。

Multi-texture Features-based Trade-mark Retrieval

A multi-texture features-based trademark retrieval is proposed.First, texture features of the image are analyzed and texture coarseness and directionality are chosen as two features on the point of visual. Then, the second order moment, entropy, contrast, and uniformity are described as four image features based on the image greyhound symbiosis matrix on the point of statistics. For the above six image features describing the image on different point, Euclidean distance is used to match the image comparability after they are normalized. By experiments, they show that the satisfactory results can be got by this technique.