Low-complexity video noise reduction in wavelet domain

Gupta, Neelesh; Swamy, M.N.S.; Plotkin, E.I.

doi:10.1109/mmsp.2004.1436537

Cited by 13 publications

(10 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The average PSNR values are computed over 40 frames for "Salesman", 52 for " Table Tennis", 52 for "Flower Garden", 60 for "Miss America" and 60 for "Hall" sequence. Experimental results show that the proposed algorithm outperforms [4], [5], [7], [8] and [9] for "Salesman", "Miss America" and "Hall" sequence both in terms of PSNR and visual quality but for " Table Tennis" and "Flower garden" sequence the proposed scheme underperforms [3] and [6] by approximately 0.07 -1 dB in terms of PSNR and shows comparable visual quality. The proposed scheme is more effective for sequences having average motion or small motion.…”

Section: Resultsmentioning

confidence: 94%

“…In Table I, we compare the proposed algorithm with the time domain filter named 3-D rational filter [5] and three existing wavelet based denoising schemes [3], [4] and [6] and in Table II we compare the proposed algorithm with another three existing wavelet based denoising schemes [7], [8] and [9]. The average PSNR values are computed over 40 frames for "Salesman", 52 for " Table Tennis", 52 for "Flower Garden", 60 for "Miss America" and 60 for "Hall" sequence.…”

Section: Resultsmentioning

confidence: 99%

“…Denoising of video sequences [2], [3], [4], [5], [6], [7], [8], [9] has attracted some attention in the past few years with different types of algorithms. Some schemes utilizes the temporal redundancy with motion estimation [1], [2], [3].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Video Noise Reduction in 3-D Mixed Transform Domain Using Its Efficient Wavelet Structure

Nath

Hazarika

Mahanta

2009

2009 Annual IEEE India Conference

View full text Add to dashboard Cite

In this paper, we propose a novel 3-D mixed transform based video denoising scheme for additive white gaussian noise. The 3-D mixed transform consists of lapped biorthogonal transform (LBT) in vertical and horizontal directions and discrete cosine transform (DCT) in temporal direction. The efficient wavelet structure of 3-D mixed transform coefficients is used for denoising the video data. The concept of wavelet structure of 3-D mixed transform coefficients is taken from our recent successful lossy color image compression scheme. The wavelet structure of 3-D mixed transform coefficients allow to use efficient wavelet based image denoising algorithms without wavelet transform. We use the approximate minimum mean square error (MMSE) estimation procedure to denoise each individual slice of 3-D mixed transform coefficients. We propose to use the simple selective recursive temporal filtering with the proposed denoising scheme which improves the average peak signal to noise ratio (PSNR) further by 1dB -2dB. Experimental results demonstrate that the proposed scheme shows encouraging results when compared to many existing complex wavelet based denoising schemes both in terms of PSNR and visual quality.

show abstract

Section: Resultsmentioning

confidence: 94%

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Video Noise Reduction in 3-D Mixed Transform Domain Using Its Efficient Wavelet Structure

Nath

Hazarika

Mahanta

2009

2009 Annual IEEE India Conference

View full text Add to dashboard Cite

show abstract

“…However, none of the above schemes explicitly exploit the Owing to spatial clustering, temporal redundancies or corretemporal redundancies that exist in the wavelet domain. In lations occur between corresponding subband coefficients in[8], we have shown that the local correlations that exist in consecutive frames in the wavelet domain representation of the temporal direction in time domain also persist in the a video sequence [8]. Even in sequences depicting motion, wavelet domain and can be exploited to develop adaptive the temporal correlations persist in the regions not affected by spatio-temporal filters.…”

mentioning

confidence: 99%

“…[8], we have shown that the local correlations that exist in consecutive frames in the wavelet domain representation of the temporal direction in time domain also persist in the a video sequence [8]. Even in sequences depicting motion, wavelet domain and can be exploited to develop adaptive the temporal correlations persist in the regions not affected by spatio-temporal filters.…”

mentioning

confidence: 99%

Video Noise Reduction in the Wavelet Domain using Temporal Decorrelation and Adaptive Thresholding

Gupta

Swamy

Plotkin

2006 IEEE International Symposium on Circuits and Systems

Self Cite

View full text Add to dashboard Cite

This paper proposes a novel wavelet domain based noise reduction scheme that is based on decorrelation of the spatio-temporal filter for video denoising that exploits the tem-temporal data, followed by adaptive thresholding. The discrete poral as well as spatial correlations which exist in the subband cosine transform (DCT) is used for temporal decorrelation representation of the video sequence. The temporal redundancy cne is usedtfor tmoral decorrlion or correlation among the corresponding wavelet coefficients inn the next section along with the outline of neighboring frames is minimized by the use of discrete cosine the algorithm. Section III describes the hierarchically adapted transform. These decorrelated noisy wavelet coefficients are thresholding scheme for spatial noise reduction. Experimental then denoised spatially via a low-complexity wavelet shrinkage results are provided in Section IV, followed by the conclusions method, which utilizes the correlation that exists between sub-in Section V. sequent resolution levels. The proposed scheme shows promising results and outperforms state-of-the-art spatio-temporal filters II TEMPORAL DECORRELATION USING DCT in time as well as wavelet domains, both in terms of PSNR and visual quality.Consider a noise-free video sequence whose nth frame can be represented as f'. Each of the noise-free frames in the original video sequence is corrupted with a set of zero-mean Digital video, which can be considered as a sequence of AWGN, q', having a variance or2, so as to yield the observed images or frames, has become vastly popular during the last frames of the noisy video sequence: decade. The corruption of such sequences by noise is also a common phenomenon, which hinders further processing of the g F + 91(1) video or its compression. Most of the existing techniques for This work is concerned with obtaining an estimate, f1, of the the removal of additive white Gaussian noise (AWGN) (e.g., noise-free frame, fn, given the observed noisy frame, gn.[1-4]) consider the video as spatio-temporal data and extendThe input noisy frame is transformed to the wavelet domain ID or 2D filters to exploit the spatial as well as temporal using the critically-sampled discrete wavelet transform (DWT) redundancies that exist in a video sequence. However, most and owing to the linearity and orthogonality of the wavelet of these schemes fail to reduce noise sufficiently.transform, the corrupting noise in the wavelet domain is also More recently, following the success of wavelets for noise additive: removal in still images, wavelet based video noise reduction yn = xn + nschemes have been proposed in [5,6].[5] uses the wavelet domain for spatial denoising only and exploits the temporal where yn and Xn are the sets of noisy and noise-free wavelet redundancy in the time domain using weighted averaging. On coefficients and n is the set of corrupting zero-mean AWGN the other hand, [6] employs weighted averaging over a spatio-coefficients having a variance r2. temporal window in the wavelet domain to...

show abstract