Is two-dimensional PCA equivalent to a special case of modular PCA?

摘要

In this note, we point out that 2D-PCA is not a special case of MPCA. 2D-PCA views the rows of images as training samples that constitute m sub-training sets instead of original images, where m is equivalent to the row of images. It then uses these sub-training sets to evaluate the covariance matrix for images. Finally, the main difference between these two methods is shown.