Fig. 9

Illustration of the alternative way of computing the projection of a tiling of Gaussian functions. In the first approach (top line), coefficients are evenly placed on the image support (left). This image is then convolved by the 2D Gaussian kernel (green circles), which gives an intermediate image (center). This image is projected to obtain a sinogram (right). In the second approach (bottom line), isolated coefficients (left) are projected. Each line of the resulting sinogram (middle) is convolved by a 1D Gaussian kernel, to obtain the sinogram (right)