基于矢量量化的颜色图像检索方法  

Color Image Retrieval Method Based on Vector Quantization

作　　者：陈善学[1] 于佳佳[1] 李俊[1] 韩勇[1] 

机构地区：[1]重庆邮电大学移动通信安全研究所,重庆400065

出　　处：《电视技术》2013年第23期35-38,45,共5页Video Engineering

基　　金：国家自然科学基金资助项目(61271260;61071116;61102062);国家科技重大专项基金资助项目(2009ZX03001-004-02);重庆市自然科学基金资助项目(CSTC2010BB2407)

摘　　要：针对颜色空间量化和直方图检索存在的问题,提出一种新的颜色图像检索方法。该方法通过对颜色空间进行矢量量化,使量化结果更接近人的感知;在提取分块主颜色的基础上引入颜色转移矩阵,以描述颜色的空间分布情况;并将索引直方图和颜色转移矩阵两种颜色特征相结合,对图像整体和局部感兴趣区域赋以不同的权值,使局部图像特征矢量增强,从而提高检索精度。实验表明,该方法能较有效地提高检索效率。A new color image retrieval method is put forward for the problems existing in color space vector and histogram retrieval. The method makes the quantized results more close to the human perception by quantizing the color space vector. The color transfer matrix is introduced on the basis of extracting the block main color aim at describing the color spatial distribution. The method combines the retrieval characteristics of index histogram with the color transfer matrix, assigns different weights between the whole region and the interested region and enhanced the partial image feature vector, thus the retrieval precision is improved. The simulation results show that this method can improve search efficiency effectively.

关 键 词：图像检索 矢量量化 感兴趣区域 索引直方图 

分 类 号：TN919.8[电子电信—通信与信息系统]