摘要
为充分发挥全局不变矩对图像形状信息着重考虑的优势,规避颜色直方图未曾考虑图像的空间位置以及常见分块策略忽略分块间的相互联系等弊端,提出一种改进的全局不变矩和分块主颜色相结合的图像检索算法。计算图像的全局不变矩,提取图像各分块主颜色,将两者进行归一化处理后进行线性加权求和,计算综合相似度。实验结果表明,该算法具有较高的检索精度,能够提高检索性能。
To give full play to the role of the global moment invariant on considering the shape information of images and to solve the problem of lacking spatial location information in the traditional color histogram and losing weight relationship between the sub-blocks in the traditional block strategy ,an algorithm of the image retrieval based on the moment invariant and the sub-block dominant color was presented .The moment invariant of the image was calculated firstly ,and each sub-block dominant color was extracted secondly ,then the similarity was measured by linear weight summation after normalization processing .Experimental results show that this algorithm can keep high retrieval precision and improve the retrieval performance .
出处
《计算机工程与设计》
CSCD
北大核心
2014年第10期3523-3526,3583,共5页
Computer Engineering and Design
基金
山西科技攻关基金项目(20090322004)
