Restoration of original image from deteriorated image by probabilistic image model
Yuji Karita, Toshiyuki Tanaka · 2008
The conventional noise removal methods are based on spatial filtering and frequency filtering. But these methods have problems associated with degradation of image along side the noise removal. In this study, we propose the method that formulates noise based on multi-dimension Gaussian distribution and restore original image from deteriorated image by Probabilistic inference based on Bayesian statistics. The effectiveness of the proposed method has been validated using benchmark images.