Wavelet-Based Video Denoising Using Gauss-Hermite Density Function

A new wavelet-domain video denoising scheme is proposed that exploits the Gauss-Hermite probability density function (p.d.f.) for spatial filtering of the noisy frame wavelet coefficients. It is observed that the proposed p.d.f. matches the empirical one very well as compared to other conventional density functions such as the generalized Gaussian and Bessel K-form densities. The proposed p.d.f. is used in an approximate minimum mean square error estimator for spatial filtering. Temporal filtering of a video sequence is performed by a motion detector and recursive time-averaging. Simulation results on standard video sequences show improved performance both in visual quality and in terms of peak signal-to-noise ratio as compared to other recent video denoising methods.