BLIND ESTIMATION OF ADDITIVE NOISE VARIANCE FOR NOISY SIGNALS IN DCT DOMAIN

Артем Юрьевич Харьков, Владимир Васильевич Лукин

Abstract


In many practical situations, signals are corrupted by noise and it is desired to apply filtering to remove it. For most of modern filters, it is needed to know noise variance a priori or to pre-estimate it in a blind manner in the presence of signal component. Requirements to methods of blind estimation are formulated and it is shown that it is difficult to satisfy these requirements. Based on the used model of additive white Gaussian model, a method
operting in DCT domain is considered and thoroughly studied. The choice of test signals is motivated. Local estimates obtained on blocks of different size are studied and it is demonstrated that these local estimates, although based on robust estimates of data scale, can be sufficiently influence by signal component that leads to a certain percentage of large amplitude DCT coefficients in data sample. It is then shown that such abnormal local estimates have to be rejected (or their influence on the final estimate should be minimized). This is done by robust processing of local estimates. It is established that block size considerably influences accuracy characterized by bias of estimates and their variance. The role of bias is dominant – noise standard deviation is overestimated - and the main task is to decrease it. According to experiments carried out for ten variants (parameter sets) of estimation method, the best results are, on the average, obtained if block size is equal to 32 and local estimates are processed using sample median. Computational efficiency is analyzed and it is shown that processing can be done quite quickly. This allows expecting real-time implementation for such applications as electrocardiogram and speech processing

Keywords


blind estimation; noise variance; DCT; robustness; signal-to-noise ratio; test signals

References


Lonera, R. L., Uilkinsona, G. N. Ustoichivye statisticheskie metody otsenki dannykh [Stable statistical methods of data evaluation]. Moscow, Mechanical engineering Publ., 1984. 230 p.

Marchuk, V. I., Shraifel', I. S. Metody vydeleniya poleznoi sostavlyayushchei pri apriornoi neopredelennosti i ogranichennom ob"eme rezul'tatov izmerenii. Monographiya [Methods for distinguishing a useful component with a priori uncertainty and a limited volume of measurement results]. Shakhty, Russia, YuRGUES Publ., 2008. 163 p.

Kornil'ev, E.A., Prokopenko, I.G., Chuprin, V.M. Ustoichivye algoritmy v avtomatizirovannykh si-stemakh obrabotki informatsii [Stable algorithms in automated systems of information processing]. Kiev, Equipment Publ., 1989. 224 p.

Khuang, T.S., Eklund, Dzh.O., Nussbaumer G.Dzh. i dr. Bystrye algoritmy v tsifrovoi obrabotke izobrazhenii [Fast algorithms in digital image processing]. Moscow, Radio and Commnications Publ., 1984. 224 p.

Abramov, S. K., Litvyak, Yu. A., Lukin, V. V., Proskurin, A.Yu. Modifitsirovannyi sigma-fil'tr dlya obra-botki odnomernykh informatsionnykh signalov [Modified sigma filter for processing one-dimensional information signals]. Sistemy obrabotki informatsii - Information processing systems, Khar'kov, NANU, PANR, KhVU, 2002, vol. 2(18), pp. 103-112.

Besedin, A. N., Zelenskii, A. A., Kulemin, G. P., Lukin, V.V. Obrabotka sluchainykh signalov i protsessov [Processing of random signals and processes]. Tutorial, Khar'kov, Nat. Aerosp. Un-ty «Khar'k. Aviats. In-t» Publ., 2005. 469 p.

Donoho, D. L., Johnstone, I. M. Adapting to unknown smoothness by wavelet shrinkage. Journal of American Statistical Association, 1995, vol. 90, no. 11, pp. 1200-1224.

Lukin, V. V., Fevralev, D. V., Abramov, S. K., Peltonen, S., Astola, J. Adaptive DCT-based 1-D filtering of Poisson and mixed Poisson and impulsive noise. Proceedings of LNLA, CD-ROM, Switzerland, August 2008. 8 p.

Nikolaev, N., Nikolov, Z., Gotchev, A., Egiazarian, K. Wavelet domain Wiener filtering for ECG denoising using improved signal estimate. Proc. IEEE Int. Conf. Acoust. Speech Signal Process, Jun. 2000, vol. 6, pp. 3578–3581.

Smital, L., Vıtek, M., Kozumplık, J., Provaznık, I. Adaptive Wavelet Wiener Filtering of ECG Signals. IEEE Transactions on Biomedical Engineering, vol. 60, no. 2, Febr. 2013, pp. 437-445.

Abramov, S. K., Zelenskii, A. A., Lukin, V. V. Avtoma-ticheskoe otsenivanie dispersii oshibok izmerenii pri obrabotke odnomernykh informatsionnykh sig-nalov [Automatic estimation of variance of measurement errors when processing one-dimensional information signals]. Collection of scientific articles “Radio-electronic and computer systems”, Khar'kov, National Aerospace University, vol. 1, 2003, pp. 5-11.

Beheshti, S., Dahleh, M. A. Noise Variance in Signal Denoising. Proceedings of ICASSP, vol. 6, 2013, pp. 185-188.

Ponomarenko, N. N., Lukin, V. V., Abramov, S. K., Egiazarian, K. O., Astola, J. T. Blind evaluation of additive noise variance in textured images by nonlinear processing of block DCT coefficients. Proceedings of International Conference "Image Processing: Algorithms and Systems II", Santa Clara, CA, USA, SPIE vol. 5014, 2003, pp. 178-189.

Abramova, V. V. A Blind Method for Additive Noise Variance Evaluation Based on Homogeneous Region Detection Using the Fourth Central Moment Analysis. Telecommunications and Radio Engineering, vol. 74, no. 18, 2015, pp. 1651-1669.

Donoho, D. L., Johnstone, I. M. Ideal spatial adaptation by wavelet shrinkage. Biometrika, vol. 81, 1994, pp. 425-455.




DOI: https://doi.org/10.32620/aktt.2018.2.06