基于有意义区域的颜色检索算法

颜色直方图方法是基于内容的图像检索系统中最常用的一种技术，然而，经典的颜色直方图方法存在诸多缺陷，例如不能表示图像中颜色的空间分布信息。为此, 提出了一种基于有意义区域的颜色检索算法，即首先计算图像的颜色聚合度，据此确定图像分割方法，对聚合度较高的图像利用区域生长方法进行分割，对聚合度较低的图像进行简单的分块，提取各区域的颜色特征并动态加权求和进行检索。该方法不仅保持了经典直方图简单方便的特点，而且有效地将空间信息集成到直方图中。在CBIR原型系统（DMIR）上的实验表明，该算法明显优于经典颜色直方图。

Meaningful regions-based algorithm for color image retrieval

An meaningful regions-based algorithm was provided: the color cluster parameters of images were calculated first.Those images with high value were split into sub-regions by image segmentation,and others were just split into sub-blocks.Each block was regarded as a region,and was given a dynamic weight.The total similarity between two images was computed according to those regions' color histograms.The experiments show that this approach can achieve increased power of discrimination than the classical color histogram method.