This page accompanies the paper "Efficient Implementation of All-Digital Interpolation"
by Bojan Vrcelj and P.P. Vaidyanathan

submitted to IP Transactions

The set of used images

images
The images are saved in the .gif format. You can also display them by clicking on the corresponding name. In each image we explicitly specified the square region used in the experiments.

Matlab code for all-FIR interpolation

fast_interpolation.tar.gz
You should be able to download this compressed file to the desired destination and then type "tar xvfz fast_interpolation.tar.gz". This should produce three .m files which are used for interpolation. Then from Matlab prompt you can type "help fast_interpol" for more information.

General comment

The code should closely correspond to the algorithm described in the paper. The resulting interpolated images will be exactly m times the size of the input regions. However, the boundary effects might affect several rows/columns of pixels next to the boundary. Therefore, these pixels were not taken into account when we computed the PSNR values reported in the paper. The number of affected pixels will in general depend on the spline order and the interpolation ratio m. The boundary effects are not only visible in this method for interpolation. Even the method we were comparing to assumes some model for the pixels next to the boundary that would often produce the result which is pleasing to the eye; however, the boundary effects are present there also. Because of all that, we compared only the stationary signal behaviour.
Go to the home pages of
Department of Electrical Engineering,
California Institute of Technology.

Maintained by Bojan Vrcelj
Last updated April, 2000