基于矩形区域重叠分块加权的图像检索

An image retrieval method based on overlapped block weighting in rectangular region

为了有效地描述图像的颜色空间分布信息并进行图像检索,提出了一种基于矩形区域重叠分块加权的图像检索方法。该方法首先裁剪待检索图片和图像库图片,得到由图像中心到外,包含小、中、大三个矩形的图像和图像库;其次分别计算三个矩形区域的彩色直方图特征及其相似度,根据空间距离度量从小到大排序;最后引入人机交互的反馈机制,根据用户的输入参数多次反馈,在用户感兴趣的图像区域检索。仿真实验结果表明,相比于均匀分块直方图算法和环形分块算法,当检索主体信息处于边缘的图像时,提出算法的图像检索率更高。提出的算法可以有效地描述图像的颜色空间分布信息并进行图像检索。

In order to describe the spatial distribution information of images effectively and to retrieve images,an image retrieval method based on overlapping overlapped block weights is proposed.In this method,retrieve images and image gallery images arefirstly cut from the centre of the image into small,medium and large outsourcing to obtain a rectangular image and image database;secondly,the colour histogram features of three rectangular regions are calculated respectively,and the similarity retrieval is done according to spatial distance measure.Finally,the feedback mechanism of human-computer interaction is introduced to retrieve the user＇s interested image region according to the feedback of the user＇s input parameters,which can greatly improve the retrieval efficiency of the image.Simulation results show that compared with the uniform block histogram and the ring block algorithm,the proposed algorithm has higher image retrieval rate when the retrieval agent is in the edge of the image.The proposed algorithm can also effectively describe the colour space distribution information of images and retrieve images.