基于有序数据聚类的图像自适应分条算法  

Adaptive image strip dividing algorithm based on ordered data clustering

作　　者：李新丰[1] 杨淑德[1] 华顺刚[1] 

机构地区：[1]大连理工大学机械工程学院,辽宁大连116024

出　　处：《计算机应用研究》2015年第3期957-960,共4页Application Research of Computers

基　　金：国家部委预研项目(40401010105)

摘　　要：为了将图像中内容特征相近的像素尽可能分割到同一区块,提高图像分割的针对性和自适应性,提出了一种基于有序数据聚类的图像自适应分条算法。该算法首先计算图像中所有像素点的梯度值,相加每列像素梯度值得到列累积能量;然后对能量数据进行加权平滑生成连续曲线,用该平滑曲线的凹凸性自适应确定图像分条总数;最后构造图像列累积能量数据的条件距离矩阵,由已确定的分条数采用系统聚类的方法实现图像分条。分条实验结果对比表明,提出的算法能根据不同图像内容自适应地进行图像条分割,且将分条结果应用于图像内容感知缩放研究中可获得满意的缩放效果,因此该算法能较好地对图像内容进行分类和识别。In order to improve the pertinence and adaptability while dividing the similar content and features within a given image into a region as well as possible, this paper proposed an adaptive image strip dividing algorithm based on the ordered da- ta clustering. First, it calculated the gradient values of all the pixels in the image and summed column by column to get the cu- mulated energy. Then, by weighted calculating the ordered column cumulated energy data to generate a smooth curve, it adap- tively determined the dividing strip number according to the number of the local peaks of the curve. Finally, it constructed the conditional distance matrix of the column energy value for dividing the image into strips by the methodology of system clustering analysis. Experiments show that it can be used to divide the image in terms of content and distribution and obtain satisfying effect in the application of content-aware image retargeting research. Therefore the proposed algorithm can classify and recog- nize the image content well.

关 键 词：自适应图像分条 有序聚类 列累积能量 梯度值 加权平滑 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]