用于图像检索的连通直方图方法

本文提出了一种用于图像检索的连通直方图（CRH：Connected-Region histogram)方法。连通直方图是图像中连通区域面积的统计函数，反应了图像中连通区域的概率分布，是图像内容的一种简洁表达。它具有灰度、旋转、平移等不变性，并可通过权函数来强调某些连通区域。另外，还可以根据连通区域的其他特征（如连通区域的形状、所处的相对位置等）来优化、改进连通直方图，提高图像的检索效率。实验证明，连通直方图具有一定的图像检索能力。

Connected-Region Histogram Method for Image Retrieval

A method of connected-region histogram (CRH) is proposed for content-based image retrieval. CRH is a statistic function of homogeneous gray connected-region (CR), revealing the probability distribution of CRs. CRH can roughly represent the image content. With a series of excellent features such as robustness to rotation, translation, illumination and so on, CRH cab be used for image retrieval. Moreover, regions of interest can be emphasized by means of the weight function. According to the local features of CR, CRH can be refined to improve the effectiveness of image retrieval. The experimental results confirm that the proposed method is promising for image retrieval.