摘要： 对纹理、形状特征提取、相关反馈算法进行了研究。利用熵进行纹理特征提取,利用具有平移、旋转不变性的离散函数的七个不变矩进行形状特征提取,利用相似性度量函数进行图像的相似度匹配,根据用户反馈调整查询对象和权重进行新的检索。实验结果表明,应用反馈算法后,可明显提高基于内容的图像检索的准确率。

关键词: 纹理, 形状, CBIR, 图像检索

Abstract: In this paper, extraction of texture and shape and algorithm of relevant feedback are designed. Entropy is used in the feature extraction of the texture, seven invariant moments with immutability of shift and rotation of discrete function is used in the feature extraction of the shape, similarity measuring function is used in image similarity match, searching object and weight are adjusted according to users feedback.Experimental results indicate that the performance of CBIR can be effectively improved after using the feedback algorithm.

中图分类号: 

• TP391.3