基于二维最小二乘回归的子空间分割
Two-dimensional least square regression based subspace segmentation

现实中有很多样本数据是二维的，且多数聚类方法需将二维样本数据向量化，从而导致二维数据的内部几何信息丢失. 针对这一问题，提出二维最小二乘回归子空间分割方法直接对二维数据进行聚类，将一维最小二乘回归子空间分割方法推广到二维，使得原始数据的结构信息得以保留. 在人脸数据集和哥伦比亚大学图像数据集上进行实验，结果表明该方法是有效的.
In reality，most of data is two-dimensional，and most of clustering methods process the data with vectorization first. This practice leads to loss internal geometry information of data. To solve this problem，a two-dimensional least square regression method based subspace segmentation is put forward for clustering on 2-dimensional data. 1-dimensional space is extended to 2-dimensional space，and it keeps the structure information of original data. Experimental results show that this method is effective on face databases and the Columbia University Image Library