融合SLIC的DCUT改进图像分割算法

Improved DCUT Image Segmentation Algorithm Based on SLIC

谱聚类DCUT算法能在任意形状的样本空间上聚类且收敛于全局最优解,但其缺点是计算相似度矩阵和特征向量的复杂度较高.为了提高了DCUT的算法速度,提出了基于SLIC的DCUT算法(SDCUT).SDCUT算法首先采用SLIC算法分割图像成超像素,再根据任意两个超像素的归一化直方图计算Pearson系数作为超像素之间的相似度,从而建立基于超像素的相似度矩阵,最后采用DCUT算法对超像素进行分类获得最终分割结果.在一系列图像上的实验结果表明,与几种经典谱聚类算法相比,本文方法的分割速度更快,且具有较好的分割效果.

Spectral DCUT algorithm can cluster samples in any form of feature space and has global optimal solution, But its disadvantage is that the complexity of computing the similarity matrix and the feature vector is higher. In order to improve the speed of DCUT algorithm, a new algorithm based on SLIC（S-DCUT）is proposed. S-DCUT algorithm firstly uses SLIC method to split the image into super pixels, secondly according to normalized histogram of super pixels, the Pearson coefficient is computated as the similarity between any two super pixels, therefore a similarity matrix based on super pixels is established, finally the DCUT algorithm is used to classify the super pixels to obtain the final segmentation results. Experimental results on a series of images show that the proposed method is faster and has good segmentation results compared with some classic spectral methods.