基于交叉分块直方图的图像检索方法

在基于内容的图像检索技术中,颜色是常用的图像特征。提出了一种基于交叉分块直方图的图像检索方法。首先,对图像进行交叉分块得到若干子图像;然后提取子图像的颜色直方图特征,构造特征向量;最后根据空间距离度量图像间的相似程度,确定检索输出结果。用交叉方式分块得到的子图像间相互交叉重叠,能够在获取图像颜色特征空间分布信息的同时保持图像目标的完整性,并有效抑制图像背景区域对检索结果的干扰。仿真实验结果表明,基于交叉分块直方图的图像检索方法具有优于基于均匀分块直方图的图像检索方法的性能,适用于对多种类型图像的检索。

An Image Retrieval Method Based on Crossed-blocking Histogram

Color is the common feature of the Content-based Image Retrieval. An image retrieval method based on crossed-blocking histogram is proposed in this paper. Firstly, blocking the image crossly to obtain several sub-images; Secondly, extracting the color histograms of the sub-images, which compose the eigenvector of the image; Thirdly, measuring the similarities between different images through the spatial distance, and the output images are then determined by the similarities. These sub-images blocked from one image are crossed and overlap and contain the spatial color feature, which can maintain the integrity of the image target and restrain the disturbance of the retrieval results brought by image background effectively. According to the simulating experiments, the image retrieval method based on crossed-blocking histogram can get the better performance than the method based on uniform-blocking histogram and retrieve multiple types of images.