Discrete 2-D Fourier Transform



next up previous
Next: Convolution and its Up: Fourier Transforms and 2-D Previous: Sampling of Images

Discrete 2-D Fourier Transform

In practice, images always have finite size and are obtained by some form of sampling of the scene either by a CCD camera or the eye. For an image array of size , the discrete Fourier transform and its inverse are given by

As in the case of 1-D signals, the transform is periodic in both . Also the inverse Fourier transform is doubly periodic in , that is to say that the left and right sides of the image have the same intensity, as do the top and bottom. When this assumption is not met, the discontinuity introduces some artifacts. There are several commonly used approaches to this problem. One of them is to use a window to modulate the border so that it drops to zero. Another is to extend the image by a border of zeros and yet a third is to add a mirror image of the last several rows of pixels to both the directions. Which one is used depends on what the qualitative nature of the artifacts introduced is like.



S Sastry
Sun May 4 22:28:26 PDT 1997