DSP Methods for Blur Reduction
Wiener Filter Performance Analysis
To test the performance of our implementation of the Wiener filter, we applied a
5 x 5 pixel Gaussian point spread function, then a 15 x 15 pixel Gaussian PSF
(blurring function, implemented in Matlab) to a sample
image, first a line drawing which we created for our group logo.
We then applied the Wiener filter (Wiener filter,
implemented in Matlab) to the blurred image, providing it with the correct PSF,
and a noise constant of 0.01, because the original image has no noticable noise
problems.
Original, Blurred, and Restored images (line drawing, 5 x 5 PSF)
Original, Blurred, and Restored images (line drawing, 15 x 15 PSF)
We also tested the Wiener filter on a photographic image. The method was as
above, with a 5 x 5 pixel Gaussian PSF.
Original, Blurred, and Restored images
Although the Wiener filter does not perfectly restore the blurred image to
match the original, it does provide a noticable improvement. Since the noise
constant is the only parameter which can be modified, and since noise is not
a concern in these images, there is no guesswork involved in applying this
filter.