基于分块主色的图象检索

Image Retrieval Based on Sub Regions and Primary Colors

传统的基于颜色直方图特征的图象检索方法简单且易于实现,但其直方图求交运算量过大,而且丢失了颜色的空间分布信息。本文给出了一种基于分块主色的图象检索方法。该方法将图象分成相等的九个子块,依次统计每个子块的颜色直方图,用户可根据各子块的的不同重要程度而改变各子块的权值,最后给出了这种基于分块主色的检索方法中图象相似度的计算方法。实验表明, 该方法一方面利用主颜色减少了直方图相交运算量;另一方面通过对图象分块保留了颜色的空间信息并能够有效地将目标从背景中区分出来,从而有效地提高了检索速度和检索精度。

It is easy to perform for traditional image retrieval based on color histogram, however, its catcmanon m color histogram intersection is too large and the space distribution information is lost. In this paper, we propose a new method for image retrieval which is based on Sub Regions and Primary Colors. The image is divided into 9 equal sub-region, and the weight value of every sub region can be changed by users according to its importance. Finally the method for calculating the similarity is provided. This method not only simplifies the calculation in histogram intersection but also remains the space distribution information in color histogram to some extent. It is proved that this method is more effective to distinguish the object from the background and improves both the retrieval speed and the retrieval precision.