摘要
图像的形状是描述图像的重要视觉和语义特征 ,可通过图像中像素点的区域分布表现出来。本文针对二值图像 ,运用中心投影变换 ,使二维的图像模式转化为基于角度的一维模式 ,该一维模式表示图像在等角距的各个方向上的投影 ,在此基础上实现二值图像的检索。实验结果表明 ,中心投影是图像的一个全局特征 ,中心投影变换能有效地刻画二值图像的形状信息 ,检索结果及排列顺序与主观判断很一致。
Shape is a very important visual and semantic feature used to depict image and it can be revealed by image pixel's regional distrubution.To binary image,this paper proposes a method that transit two-dimenional image pattern into one-dimensional pattern based on angle through using central projection transformation,the one-dimensional pattern is the projection of image pixels in various angles or directions,afterward,decreasing the dimension of feature space by grouping and uniting the projection datum.After completing the central projection transformation for different images,Euclidean distance is used as similar mesure between two images,the smaller the distance,the more similar the two images.The experiment results show that central projection transformation can depict image's shape feature well,when it is used in binary image retrieval,the retrieval results and the arrangement order is fit with human's subjected decision well,it can be concluded that central projection is a global feature of image and central projection transformation can extract binary image shape feature effectively.
出处
《计算机应用与软件》
CSCD
北大核心
2005年第2期21-22,75,共3页
Computer Applications and Software
基金
广东省自然科学基金资助项目(编号 :0 1 1 62 8)
