基于空间分布的图像检索算法研究

Image retrieval based on spatial distribution

针对利用颜色直方图进行图像检索时存在的问题,提出了一种基于空间分布特征的图像检索算法。首先将图像分解为8个位平面,并采用其中对表征图像结构特征有意义的4个位平面的分布特征对图像内容进行描述,同时,为避免图像中像素灰度值的微小变化对位平面的影响,又提出采用位平面的格林码表示方法。考虑到位平面之间的相关性,设计了权值矩阵,采用马氏距离进行相似性度量。实验结果表明,该算法利用位平面分解的方法来提取颜色的空间分布信息,可避免颜色量化引起的误检问题,而且提取的特征矢量维数较低,具有较高的检索率。

Based on the analysis of the problems in image retrieval using color histograms, this paper proposes a novel spatial distribution-based method for image retrieval. The method firstly divides the image into eight bit-planes and introduces the Gray code to avoid the effect of the variation of their intensity values. Then, a feature vector, including the distribution information of the first four significant planes, is presented as an index for image retrieval. It combines with the new designed correlation-weighted matrix, and adopts the Mahalanobis distance as a measure rule. The comparisons with other related descriptors were conducted and the experimental results showed that the proposed method provided more significantly results than the traditional ones.