基于核主成分分析和子空间分类的边缘检测方法

Edge Detection Method Based on Kernel Principal Component Analysis and Subspace Classification

针对传统边缘检测方法对噪声敏感的问题,提出了一种基于核主成分分析和子空间分类的边缘检测方法,建立了统一的图像特征表达模型.首先结合其它边缘检测方法进行采样并将采样结果投影到特征空间,然后将核主成分分析得到的特征向量组成特征空间的一个子空间,最后将子空间分类法推广到特征空间来对数据进行分类.实验结果表明,该方法增强了对噪声的鲁棒性,能适应小样本训练,其边缘检测效果明显优于经典算子、主成分分析和非线性主成分分析方法.

In order to enhance the robustness of the traditional edge detection methods to noises, an edge detection method based on the kernel principal component analysis （KPCA） and the subspace classification is proposed, and a unified model to represent image features is established. First, the proposed method combined with other edge detection methods selects samples which map in the feature space, and then builds a subspace in the feature space with the eigenvetors obtained via KPCA..Afterwards, it expands the subspace classification into the feature space for data classification. Experimental results indicate that the proposed method is robust to noises and is suitable for small-sample training, and that the detection accuracy of the method is higher than that of the classical operators, the principal component analysis （PCA） and the nonlinear PCA.