Artificial image containing a few bars, lines and a square. The intensity is constant along each individual bar; from left to right, the intensities of the ten vertical bars (these are in fact thin rectangles which are 4 pixels wide and 170 pixels long) are equal to 32 / 2^i, i = 0,... 9. The intensity along all the other lines is equal to 1, and the noise standard deviation is 1/2. All images are log transformed (t = log(ima - min(ima) +1)), in order to better see the behavior of the methods at low signal to noise ratio. Right, Simulated image + Gaussian noise (sigma=0.5).

Filtered image by the undecimated wavelet transform (6 scales and 4sigma detection) and filtered image by the curvelet transform (6 scales and 4sigma detection).

Left,filtered image by the curvelet transform (6 scales and 4sigma detection), and using three spanning cycles (shifts = (0,0), (0,8), and (8,0). Right, filtered image by the curvelet transform (6 scales and 4sigma detection), and using three spanning cycles (shifts = (0,0), (0,16), and (16,0).

Simulated image + Gaussian noise (sigma=0.4), and filtered image by the undecimated wavelet transform (6 scales and 4sigma detection).

Filtered image by the curvelet transform (6 scales and 4sigma detection).