基于颜色特征和空间特征的图像检索

Image Retrieval based on Color Features and Spatial Features

针对已有的综合多特征进行图像检索的算法复杂度高、计算量大且准确率不够的现状,提出一种基于图像分块并综合灰度特征和空间特征的图像检索算法。首先将原图像划分为10×10的子块,再计算每个子块的平均灰度和空间位置作为检索特征,通过实验表明这种方法在检索结构相似的图像时准确率高,但对颜色不敏感。通过增加颜色特征并调整空间特征计算方式进行改进,相比改进前,算法对于公开的图像库中大多图片有更高的查准率。实验中统计了多类图片在不同情况下的查准率,表明文中算法准确高效,应用性好。

The existing multiple features based image retrieval algorithms are complex,which need large amount of calculation and are lack of accuracy.An image retrieval algorithm based on image block and gray level features and spatial features was proposed.Firstly the original image is divided into 10 x 10 sub-block,then the average gray value and spatial position of each sub-block each is calculated for image retrieval.experimentsshow that this method has high accuracy for images with similar spatial structure,but it is not sensitive to color.By adding color features and adjusting the spatial feature calculation,the method has higher accuracy for most images of the public image library.Accuracy rates were calculated for different kinds of images in different cases by experiments,and the results show that this algorithm is accurate,efficient and easy for application.