Hossein Rabbani - Bioelectric Department, Faculty of Biomedical Engineering, Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran
Mansur Vafadust - Bioelectric Department, Faculty of Biomedical Engineering, Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran

In this paper we present a new video denoising algorithm that model distribution of wavelet coefficients in each subband with a Gaussian probability density function (pdf) that its variance is local (It means that we use a separate Gaussian pdf for each pixel of each subband). This pdf is capable of modeling the heavy-tailed nature of wavelet coefficients and the empirically observed correlation between the coefficient amplitudes. Within this framework, we describe a novel method for video denoising based on designing a maximum a posteriori (MAP) estimator, which relies on the zero-mean Gaussian random variables with high local correlation. Because separate 3-D transforms, such as ordinary 3-D wavelet transforms, have visual artifacts that reduce their performance in applications, we perform our algorithm in 3-D complex wavelet transform. This non-separable and oriented transform produces a motion-based multiscale decomposition for video that isolates motion along different directions in its subbands and prevents from directions mixing that appear in subbands of 3-D ordinary wavelet transform. In addition, we use our denoising algorithm in 2-D complex wavelet transform, where the 2-D transform is applied to each frame individually. Although our method is simple in its implementation, our denoising results achieve better performance than several methods visually and regarding peak signal-to-noise ratio (PSNR).

Complex wavelet transforms, video denoising, parameter estimation, statistical modeling