Adaptive fuzzy filter algorithm for real-time video denoising

Abstract

In order to improve the perceptual quality and the compression efficiency in video transition, a new adaptive fuzzy filter algorithm for real-time Gaussian noise removal is proposed. Based on the analysis of noise and video properties, it exploits the spatiotemporal neighbor information through noise estimation, motion detection, and spatiotemporal neighborspsila similarity measurement. Then, an adaptive fuzzy filter is applied to remove noise. Experimental results showed that the proposed method achieved better performances of noise removal and detail preservation. Compared with other methods, PSNR of the denoised video was improved 1~ 6 dB. And, benefited from its low complexity and fast computation, it could meet the real-time requirements of the video transition.