中文说明：Wright 等人[13]最近几年研究的一种从低秩矩阵恢复问题中引导出的Robust PCA，引起了很多关注，也是目前最为流行的RPCA 方法。低秩矩阵恢复本义是从带有噪声的数据中恢复出原始的低秩数据，可以看到其思想与PCA 是类似的，因为PCA 是要找到数据的低维子空间，数据中不属于低维子空间的部分即为矩阵恢复中的噪声。其具体思想为：在(1)中，不仅仅要求L0是低秩的，还要求S0是稀疏的，并且S0中的元素可以是任意大的。通过这样的假设，即使数据中存在野点，也就是个别像素的噪声十分大，RPCA 也能够将这个噪声划分到稀疏的矩阵中去，如果能够很好地解这个优化问题，那么RPCA 问题就得到了很好的一个解。
English Description：
In recent years, Wright et al. [13] studied a kind of robust PCA guided from the low rank matrix restoration problem, which has attracted a lot of attention, and is also the most popular RPCA method at present. The original meaning of low rank matrix recovery is to recover the original low rank data from the data with noise. We can see that its idea is similar to PCA, because PCA is to find the low dimensional subspace of the data, and the part of the data that does not belong to the low dimensional subspace is the noise in matrix recovery. The specific idea is: in (1), not only l0 is required to be low rank, but S0 is also required to be sparse, and the elements in S0 can be arbitrarily large. Through this assumption, even if there are outliers in the data, that is, the noise of individual pixels is very large, RPCA can also divide the noise into sparse matrix. If the optimization problem can be solved well, then RPCA problem will get a good solution.