基于位平面分布特征的图像检索算法

Image Retrieval Based on Bit-plane Distribution Feature

针对利用颜色直方图检索时存在的问题,提出了一种基于位平面空间分布特征的图像检索算法.首先结合光照、锐化、模糊等噪声的攻击特点,将图像分解为8个位平面,然后提取对表征图像结构特征有意义的4个位平面的空间分布信息组成特征矢量来多层次地对图像特征进行描述.为了避免图像中像素灰度值的微小变化对位平面的影响,又提出了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,最后设计了权值矩阵,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特征,又避免了颜色量化引起的误检问题,而且提取的特征矢量维数较低,因而具有较高的检索率.

The contentbased image retrieval on histogram was analyzed and a novel image retrieval algorithm was proposed based on bitplane distribution feature. Firstly, the feature of the noise was considered and the image is divided into eight bitplanes by the image bitplanecode. Then, the Bitplane Distribution Feature （BPDF） was constructed according to statistic distribution of the four significant planes which contain most of the structural information of the image. Meantime, the graycode of bitplanes was used to avoid the effect of changes in the image intensity values on bitplanes. Finally the Mahalanobis distance was adopted to measure the similarity because of the correlation between the concerned vectors after designing the correlationweighted matrix. Experimental results show that the proposed method has sound retrieval performance.