Figure 8
From: An augmented Lagrangian multi-scale dictionary learning algorithm
An illustration of overlapped patches' distributions from different images. Top: the standard covariance matrix of the 62001 examples extracted from "Barbara" (a) and "Cameraman" (b), and σ = 20. Each entry of the matrix is between 0 (black) and 1 (white). Bottom: The first two dimension projection of the 62001 examples extracted from "Barbara" (c) and "Cameraman" (d), and σ = 20.