序列化颜色直方图的分析与应用

在许多基于颜色的图像检索系统中,图像的颜色属性通过颜色直方图来描述。基于颜色直方图的搜索,对规模较大的颜色直方图往往很低效。因此论文将介绍一种新方法,这种方法能有效的表示颜色直方图和两个颜色直方图的区别。这种方法是基于以下的观察:最大的20%的颜色直方图能占到图像的90%的像素;当颜色直方图用一个阀值过滤后,主成分颜色数量MCCC(maincomponentcolorusedcount)代表这幅图像的复杂性,在第一阶段的检索中被用来快速索引,主成分颜色序列MCCS(maincomponentcolorsequence)能够用来在检索的第二阶段比较图像的颜色信息。使用该方法,总的颜色索引信息将比传统的颜色直方图节省存储空间。试验结果基于包含1000张图像的图像数据库。

The Analysis and Applications of Ordered Color Histograms Sequence

In many color-based image retrieval systems the color properties of an image are described by its color histogram.Histogram-based search is,however,often inefficient for large histogram sizes.Therefore we introduce a new method that provides efficient representations of color histograms and difference between two color histograms.The method is based on the following observation:the max 20% color bins in the color histogram account for about 90% pixels in the image.When the color histograms are filtered by a threshold,the main component color used count,MCCC,representing the complexity of image,can be used for fast indexing at the first retrieval stage,and the main component color sequence,MCCS,can be used for comparing images about color information at the second stage.Using our method,the total color index information needs less storage space compared with the traditional color histogram.In this paper we report the results of our experiments that are done on the image database containing 1,000 images.