Image restoration is a process to reconstruct or recover an image that has been degraded by using some a prior knowledge of the degradation phenomenon. We consider the degradation is composed of random noise and spatially invariant blurring.
Most techniques for image restoration were derived from frequency domain concepts. Particularly, Wiener filtering is a well-known signal processing technique to restoration, but it needs large storage and computation time.
In this thesis, the scalar Wiener filter is derived for several unitary transforms. The performance of the filter on FFT, Hadamard, and Discrete cosine tranform is compared. The image is assumed degraded by additive white noise and uniform motion blur, or Gaussian blur.
