Demonstration of compressive sensing as applied to an MR image. A, B and C are the log of the Fourier domain, D, E and F are the images and G, H and I are the log of the wavelet domains. Column 1 is the original image, column 2 is the image undersampled by a factor of 5 (1/5 of the points in the Fourier domain were used in a random, centre-weighted Gaussian pattern), and column 3 is the CS reconstructed image. This is a screenshot from a real-time demo optimized for visual impact. The algorithm used is a simple (but fast) thresholding approach and does not perform as well as more sophisticated algorithms.