【作者】 洪清启； 王备战； 董槐林； 张磊； 陈兵；

【Author】 HONG Qing-qi,WANG Bei-zhan,DONG Huai-lin,ZHANG Lei,CHEN Bing (School of Software,Xiamen Univ.,Xiamen 361005,China)

【机构】 厦门大学软件学院；

【摘要】 传统的基于颜色直方图特征的图像检索方法简单、高效,但是其完全丢失颜色的空间分布信息;针对这个缺点,有人提出了基于广义直方图的图像检索方法,然而其也没有考虑到像素间的方向信息,丢失了图像的局部细节信息。在广义直方图的基础上提出了改进的基于广义直方图的检索方法,新方法首先计算出原图像在0°,45°,90°,135°4个方向上的平滑图像,并将其分别与原图像对应的象索点组成二元组,统计每个二元组出现的概率,构造出4个方向的广义直方图,最后综合4个广义直方图进行检索。实验表明,与原始的广义直方图法相比,新方法能够较准确地描述出图像象素间的空间位置关系及方向细节信息,从而有效地提高了检索的精度。更多还原

【Abstract】 The color histogram based image retrieval method is simple and efficient,however,it has completely lost the information of color spatial distribution;in allusion to this disadvantage,the retrieval method based on generalized histogram was proposed,but it did not consider the orientation info between border upon pixels and lost the detail info of the image.This paper proposes an improved method based on generalized histogram.Firstly,it figures out the four smoothed images of the original image in the four directions of 0°,45°,90°,135°.Secondly,it combines the original image with one of the four smoothed images to form a binary set,which is just the generalized histogram. There are four smoothed images in the four directions,so it constructs four generalized histograms in the four directions respectively.Finally,it uses the four generalized histograms for retrieval.The experimental results show that compared with the original generalized histogram,the new method could describe the info of color spatial distribution in the picture and orientation info between pixels more properly and improves the image retrieval precision.更多还原