Abstract: In the view of image retrieval, this paper proposes a new block weighted image retrieval method based on visual features. According to visual attention mechanism, this paper divides the image into several non-uniform blocks, and sets different weights to extract visual features of each block. Firstly, Sobel operator is used to obtain the edge of the image to calculate the color difference histogram of color and edge orientation. Meanwhile, this paper defines a structure which combines color and edge orientation and uses histogram to represent it. Moreover, the histogram of each block is connected to generate the feature vector of an image for retrieval. This paper carries out the experiments on benchmark Corel image database, and the extensive experiments demonstrate that the proposed method achieves better retrieval performance in comparison with state of the art image retrieval algorithms.

Key words: image retrieval, block weighted, visual attention mechanism, Sobel operator, color difference histogram

摘要： 对图像检索进行了研究，提出了一种新的分块加权图像检索方法。根据视觉注意机制将图像分成不均匀的若干块以及设置不同的权值，提取每块的视觉特征。首先在HSV颜色空间利用Sobel算子得到图像的边缘信息，计算颜色和边缘方向的色差直方图，同时定义一种结合颜色和边缘方向的结构来获取图像的纹理信息，并用直方图表示；然后连接每个块的直方图作为图像的特征向量进行图像检索。实验选取Corel标准图像库进行检索以及和另外5种图像检索方法进行对比分析，实验结果表明该方法具有较高的检索精确度。

关键词: 图像检索, 分块加权, 视觉注意机制, Sobel算子, 色差直方图