基于距离分布信息熵的商标图像检索

Binary trademark image retrieval using distance distribution information entropy

提出了一种基于距离分布信息熵的图像检索方法,该方法首先对图像的目标区域进行区域划分,然后提取区域的信息熵作为特征来描述图像形状,最后使用欧式距离度量熵矢量之间的相似性。实验结果表明,距离分布信息熵能有效地刻画出二值图象的形状特征,并且具有良好的平移、旋转及尺度不变性,检索结果符合人眼的视觉感受。

A new trademark image retrieval method based on the distance distribution information entropy is presented in this paper.The proposed method is divided into the following three steps：conducting the region partition on the object of the image, computing the information entropy of each of the partitioned regions,which are constructed into a feature vector for describing the shape of the image,and finally,measuring the similarity between the images using the euclidean distance.Experimental results on the trademark image database show that the extracted features can describe the image shape better,and have good invariant property under translation,scale and rotation.