摘要
提出了一种基于多重分形分析的图象边缘提取算法 ,通过计算每个象素点的奇异值和多重分形谱 ,并根据多重分形谱的各种测度修正 ,提取出图象的边缘信息 .在分析图象的各种象素点的多重分形谱特性的基础上 ,着重分析了多重分形谱常用的若干测度以及选取标准 .该算法利用奇异性H¨older指数和多重分形谱以及它们组成的判据来进行图象边缘提取的思路不同于传统的基于梯度的局部极值点来进行图象边缘提取的方法 .实验结果表明 :该算法可以在保留重要边缘信息的情况下去除不重要细节 ,更能符合人的视觉心理 .
A multifractal analysis approach to the problem of image edge detection is proposed in this paper. Based on the value of each pixel′s singularity and its relative height, which is given by computing spectra associated with different kinds of capacities defined from grey levels, edge information of image is gotten according to the modified multifractal spectrum. The algorithm divided into three steps: firstly each point′s singularity H lder exponent α is being computed; secondly according to measure which you choose, the multifractal analysis spectrum f(α) is being computed. And this f(α) is on behalf of the whole image′s singularity. At last, the image is dectected according to (α,f(α)), the deciding rule is : if α(x,y)≤1.2 and f(α) close to 1, the point (x,y) is seemed as a edge point; and if α(x,y)≤1.2 and f(α) close to 2, the point (x,y) is seemed as a texture point. Our experiment show that in several cases, the approach gives at least as good results as the classical ones, it′s effective for edge detection and giving the eminent and detailed edge information of the image. Much more work is needed in this direction, but these preliminary results show that the (α,f(α)) approach might be able to build a bridge between two so far unconnected methods of edge detection and region extraction.
出处
《光子学报》
EI
CAS
CSCD
北大核心
2003年第1期61-64,共4页
Acta Photonica Sinica