摘要
在数字图像领域,图像的特征检测是一种重要的图像预处理技术,广泛应用于轮廓抽取和纹理分析等领域.本文在Matlab编程环境下,将脊波变换提取图像直线特怔的实现分为3个步骤:(1)对含躁声的图像进行Randon变换;(2)对得到的Randon变换域进行3层小波变换,得到脊波系数;(3)对脊波系数进行阈值处理,得到的稀疏脊波系数经逆变换提取图像的直线特性.利用该方法分别对几种含躁声的图像进行直线检测,结果图像的信噪比可以达到17.4以上.研究表明,脊波变换对直线特征的提取可以得到良好的效果.
In the area of digital image, the feature detection of an image is an important preprocessing technology which is widely used in the fields of contour and texture analysis. In this paper, based on the Matlab programming environment, the beeline feature detect from the noise image using the ridgelet transform was realized by three steps： 1） the Randon transformation was used to the noised image; 2） the three tiers of wavelet transform was used to the domain of ridgelet transform coefficient; 3） the ridgelet coefficient was thresholds and the sparse ridgelet coefficient was given, its result can be extract the beeline feature from the image using inverse transform. The method was used to detect beeline feature from the noise images, and the PSNR is large than 17.4. The research show that the ridgelet transform can take out beeline feature from the noise image and the result is favorable.
出处
《厦门大学学报（自然科学版）》
CAS
CSCD
北大核心
2006年第6期775-778,共4页
Journal of Xiamen University：Natural Science
作者简介
潘伟（1958-），男。副教授．Email：wpan@xmu．edu．cn
