An Adaptive Method for Video Denoising Based on the Ici Rule

This paper presents an adaptive video denoising technique based on the intersection of confidence intervals (ICI) rule used for adaptive filter support size calculation. The method is applied to three real-life video signals and its denoising performance is compared to a fixed size filter support based method resulting in a significant estimation error reduction in terms of the average frame peak signal-to-noise ratio (PSNR) improvement. The average frame PSNR obtained by using the here presented ICI based video denoising method is increased by up to 14.64 dB and by up to 23.74 dB when compared to the fixed size filter support based method. Furthermore, unlike the fixed size filter support based method, the adaptive ICI based method is shown to be efficient in a moving object edge preserving, while avoiding its blurring. The method performs well for both video signals obtained by recording stationary scenes, and video signals of moving objects, which are far more often encountered in practical applications, whereas the fixed size filter support based method is limited only to video signals of stationary scenes.