Suppose we have an image and we store its pixel array as a matrix. Then, we compute the SVD of the matrix. What do $U$ and $V$ the two bases qualitatively signify?
Please let me know if I am being vague. I understand that you can approximate the image by using the rank approximation theorem, but I am confused as to what the columns of $U$ and $V$ signify. Does one correspond to space (i.e. location of the pixel) and the other correspond to the color of that pixel?