摘要:
针对当前彩色图像检索技术容易受到色彩干扰,鲁棒性不强等不足,提出了一种字典统计耦合归一化多重距离的彩色图像检索算法.首先,将图像量化并转换成一维信号;然后,引入字典统计,将一维信号进行字典编码,并计算编码后的图像多样值;在归一化字典距离的基础上,嵌入字典编码图像的多样值,从而定义了归一化多重距离(NMD)的相似度量准则,利用NMD对查询图像与数据库图像的多样值进行比较与识别,搜索出与查询图像具有相同特征的最相似图像,完成目标检索.在COREL数据库的实验结果表明:相对于当前常用的检索技术,该检索算法具有更高的查准率和查全率,可对彩色图像完成精确检索,有效减低了色彩对检索性能的干扰.
Abstract:
In view of the current color image retrieval technology weakness of vulnerable to color interference,retrieval rate was not high and robust character was insufficient and so on,an image retrieval algorithm has been proposed on the basis of dictionary statistics and normalized multiple distance.Firstly,feature extraction was performed on the image,image quantization and converted into a one-dimensional signal string;then with the dictionary to coding image visual pattern,the characteristic value of the image were calculated.At last,on the basis of normalized dictionary distance,the similarity measure criterion of normalized multiple distance (NMD) was defined,with NMD to comparison and identification the characteristic value of the query image and the database image,retrieval and query images similar to the contents of the image,completing the retrieval task.Experimental results on the COREL database show that,relative to the current commonly used retrieval techniques,the proposed retrieval algorithm has better precision and recall,the average retrieval average precision and the average recall at the number of k=10 and k=100 was better than other methods,reducing the color of the performance of the interference of the retrieval,it can be effective for color image retrieval.
下载: