一种基于最大相似子阵的彩色图像检索算法  

Color Image Retrieval Algorithm Based on the Maximum Similarity Submatrix

作　　者：何国斌[1,2] 代秀娟[3] 李航[2] 

机构地区：[1]西南大学计算机与信息科学学院,重庆400715 [2]西南大学资源环境学院,重庆400715 [3]重庆工商大学计算机与信息工程学院,重庆400067

出　　处：《计算机科学》2012年第7期282-286,共5页Computer Science

基　　金：重庆市科委基金项目(2011GGB018)资助

摘　　要：颜色是图像的重要特征之一,被广泛用于图像检索。传统的颜色直方图检索算法丢弃了色彩的空间位置信息,影响了算法的检索精度。针对这一问题,提出了一种基于最大相似子阵的彩色图像检索算法,即采用奇数方阵分割图像以捕捉颜色的空间分布特征,从而得到颜色矩阵,通过求解两个颜色矩阵的最大相似子阵实现相似度计算。研究结果表明,所提算法能够准确描述图像的颜色特征,相似度计算非常有效,相比全局直方图有较好的检索精度。Color is one of the important characteristic of image, which is widely used in image retrieval. In the convew tional color histogram retrieval algorithm, the color＇s spatial location information is discarded, and the retrieval precision of the algorithm is also affected. Aiming at this problem, this paper presented a color image retrieval algorithm based on the maximum similarity submatrix. This algorithm uses the odd square matrix in image segmentation to capture the color spatial distribution pattern and get the color matrix, then,find the maximum similarity submatrix of the two color ma trix to achieve the similarity calculation. The results show that the algorithm in this paper can accurately describe the image color characteristic, and similarity calculation is very effective. Compared with global histogram, it has better retrieval precision.

关 键 词：图像检索 颜色直方图 色彩空间 颜色矩阵 相似子阵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]