The problems encountered in spatially correlated noise variance estimation in spectral domain are considered. The possibility of accuracy improvement of spatially correlated noise variance estimation in spectral domain by means of original noisy image decimation is studied. The comparative analysis of estimation accuracy while processing original and decimated test images corrupted by the spatially correlated noise for several best methods operating in spectral domain is carried out. The results of numerical simulations for different test images from TID2008 database are presented.
additive noise, blind variance evaluation, spatially correlated noise
“Problemy otsenivaniia dispersii prostranstvenno-korrelirovannogo shuma v spektralnoi oblasti i odin iz vozmozhnykh putei ikh resheniia”,
Information Processing Systems,