Image Restoration Using a Nonstationary Image Model
Long-Wen Chang and Jien-Hua Huang
Institute of Computer and Decision Sciences,
National Tsing Hua University,
Hsinchu, Taiwan (30043), ROC
This thesis is concered with improving the fast nonrecursive algorithms for the Wiener restoration of degraded images. The degradation is assumed to be a known space-invariant point spread function and an additive white noise. An improvement is achieved by modelling the image with nonstationary model, which is more appropriate for real images than the conventional stationary model. The algorithm employs a nonstationary-mean stationary-autocovariance (NMSA) image model. The root-mean-squared error (rmse) yielded is lower than that of the wide-sense stationary image model. The specific structure of the underlying model enables the implementation of the filters using fast Fourier transform (FFT) computations.