基于局部主色和边缘特征的图像检索

Image retrieval based on combining local dominant color and edge histogram

在基于内容图像检索技术中,经典颜色直方图方法具有一定的局限性,为此提出一种综合图像局部主色和边缘特征的图像检索方法.通过使用HSV颜色模型对图像进行量化,设计了一种重叠方式的图像分块方法,然后加权各分块区域主色,得到颜色特征,再利用Sobel算子提取图像边缘,用二维边缘直方图描述边缘特征,最后将颜色特征和边缘特征通过加权处理,得到图像总的相似度,实现图像检索.实验表明:综合两种特征的检索方法比单一特征的检索方法效果更好.

The use of color histogram is now a classical method in the CBIR, which has proved to be limited. An image retrieval method based on local primary color-edge multi-features is proposed. In the field of retrievaling based on color, we use HSV color model to quantify the color and design 25 overlapped sub-regions to get color histogram in each region. Image edge is extracted by using Sobel detection operator. 2D edge histogram is used to describe shape feature. By assigning a weighting value, the technology based on color and edge features combine to a multimethod. The result shows that the retrieval results obtained from combined-features are superior to that obtained from single-features.