Image restoration based on an anisotropic noncausal stochastic model
Yasuo Yoshida, Hisanao Ogura · 2005
This paper presents a line-by-line recursive filtering method to restore a monochromatic image corrupted by an additive white noise. The signal image is considered an anisotropic homogeneous random field that satisfies a noncausal stochastic difference equation driven by a white noise field. The model equation has three parameters corresponding to the maximum and minimum correlation lengths and their directions. The filtering of noncausal model is carried out by means of a hybrid algorithm using two one-dimensional techniques: namely, the fast Fourier transform and the recursive Kalman filtering. The algorithm is applied to a synthetic and a real images to check its validity.