基于颜色的图像检索因对图像的各种变化有很好的鲁棒性而得到了广泛的应用,但是由于缺乏空间信息而造成检索误差.针对全局颜色直方图和分块颜色直方图的检索问题,提出了一种利用等面积的矩形环来提取颜色特征的图像检索方法.该方法首先利用等面积的矩形环划分策略对图像进行分块；其次,提取子块的颜色累加直方图作为颜色特征；然后,为了突出图像的主体区域,按照矩形环由里到外依次减小规则确定权值,并对两幅图像子块颜色特征之间的距离进行加权累加得到两幅图像的相似度量,以此进行检索,并输出相应的查询结果；最后通过实验结果表明:该方法与全局直方图、累加直方图及传统分块直方图相比,具有更高的检索效果.%Color-based image retrieval has been widely used due to good robustness to various changes of images. However, lack of color spatial information leads to retrieval errors. According to retrieval problems caused by global color histogram and block histogram, the paper proposed a new image retrieval method that uses equal-area rectangle ring to extract image features. Firstly, we made use of equal-area rectangle ring partitioning strategy to block the image. Se-condly,we extracted sub-block accumulative color histogram as color features. Thirdly,we set weight values according to gradual reducing of rectangle ring from inside to outside in order to highlight the main part of an image, and got the similarity measures of two images by calculating weighed accumulative distance which is between the sub-blocks color features of two images, then retrieved images and outputed the retrieval results. At last,experimental results show that the proposed method has a higher retrieval effect, comparing to global histogram, cumulative histogram and traditional block histogram.
展开▼
