摘要
针对目前基于颜色的图像检索算法在颜色特征提取的不足,提出一种新的颜色特征提取算法。利用LBG算法对HSI空间的颜色信息矢量量化,然后统计图像中各个码字出现的频数,形成颜色直方图。这样在提取颜色特征过程中,尽可能地降低图像原有特征失真。同时通过设定门限值,多次实验比较查全率和查准率,找到较为满意的门限值,使检索算法更加完善。实验结果表明,该算法能有效地提高图像检索精准度。
We put forward a new colour feature extraction algorithm for the shortcoming of present colour-based image retrieval algorithm in colour feature extraction. First,the algorithm uses LBG algorithm to carry out vector quantification on colour information in HSI space,and then counts the appearance frequency of each code word in the image to form colour histogram. So in the process of colour feature extraction the distortion of original image features can be reduced as far as possible. Meanwhile,by setting the threshold value we compared the recall and precision rates through a couple of the experiments until a satisfied threshold value was found,thus made the retrieval method more perfect. Experimental results showed that the new algorithm could effectively improve the accuracy of image retrieval.
出处
《计算机应用与软件》
CSCD
2016年第3期222-225,共4页
Computer Applications and Software
关键词
HSI空间
图像特征
矢量量化
门限值
HSI space
Image feature
Vector quantification
Threshold value
