Dmytro Rusanovskyy and Karen Egiazarian, ACIVS 2005, Antwerp, Belgium Video Denoising Algorithm in Sliding 3D DCT domain Dmytro Rusanovskyy and Karen Egiazarian Institute of Signal Processing Tampere University of Technology, Finland E-mails: {FirstName.LastName}@tut.fi Abstract. The problem of denoising of video signals corrupted by additive Gaussian noise is considered in this paper. A novel 3D DCT-based video- denoising algorithm is proposed. Video data are locally filtered in sliding/running 3D windows (arrays) consisting of highly correlated spatial layers taken from consecutive frames of video. Their selection is done by the use of a block matching or similar techniques. Denoising in local windows is performed by a hard thresholding of 3D DCT coefficients of each 3D array. Final estimates of reconstructed pixels are obtained by a weighted average of the local estimates from all overlapping windows. Experimental results show that the proposed algorithm provides a competitive performance with state-of- the-art video denoising methods both in terms of PSNR and visual quality. 1. Introduction Digital images and video nowadays are essential part of everyday life. Often imperfect instruments of data acquisition process, natural phenomena, transmission errors and compression can degrade a quality of collected data. Presence of noise may sufficiently affect the further data processing such as analysis, segmentation, classification and indexing. Denoising is typically applied before any aforementioned image/video data processing. Herein, the problem of denoising of video corrupted by additive independent white Gaussian noise is considered. Historically, first algorithms for video denoising operated in spatial or spatio-temporal domains [1]. Recent research on denoising has demonstrated a trend towards transform-based processing techniques. Processing in a transform domain (e.g. in DCT, DFT or wavelet domains) provides a superior performance comparing to the spatio-temporal methods due to a good decorrelation and compaction properties of transforms. Wavelet-based video denoising was inspired by the results of the intensive work on the wavelet-based image denoising [3-5] initiated by Donoho’s wavelet shrinkage approach [2]. Several multiresolution (wavelet-based) approaches were recently proposed to the problem of video denoising, see, e.g. [6] and [7]. Local adaptive sliding window DCT (SWDCT) image denoising method [8], [9] is a strong alternative to the wavelet-based methods. This paper gives an extension of it to SWDCT denoising of video. This extension is not a straightforward one. Video data in the temporal direction are not stationary due to a motion present in videos. Thus,