Probabilistic Nonlocal Means Image De-noising in Tele-radiology

Author Name(s): J.SaiKrishna* M.Venkatanaresh P.V.S.R Bharadwaja .Dushyanth4
Author Email: saikrishna.j@svcolleges.edu.in

Abstract

Tele-radiology   is   one   among   the   digital medical diagnosis systems, where the transmission and maintenance of diagnosed Medical reports like X-Ray, C.T Scan and MRI, are accessed through internet. While transmitting such medical reports in the channel it may be affected by Noise at different sources. Restoring the original image from noisy image is a critical issue, where   Nonlocal means is one of Spatial de-noising techniques. In this technique, de-noising is initiated from the center patch by using weighted versions of all patches in search neighborhood. Process of Searching neighborhood includes some dissimilar patches. We use hard thresholding based on distribution of distances of similar patches that eliminates the  dissimilar  patches.  Probabilistic  Non  Local  Means P-NLM denoted by Elimination of Dissimilar Patches in Nonlocal Means (NLM-EDP).This NLM-EDP method improve PSNR and SSIM of the retrieved image in comparison with nonlocal means.

Introduction

Tele-radiology comes under the stream of electronic health care, where monitoring and diagnosing of patients is done through internet. It is more concern about paperless, wireless accessing of medical reports like MRI, C.T. Scan. Transmitting medical reports through wireless channels may be effected by noise, where restoring the original image from noisy image is a critical issue. From few decades we use to some de-noising techniques by using mean, median, Gaussian, and  bilateral  filters  [1].

Conclusion

De-noising   the   Bio-medical   images   through   additional pre-processing hard thresholding before PNLM is implemented on noisy images and observed de-noised images. The same   de-noising is replicated with conventional de-noising schemes NLM, PNLM, but significant improvement in all metrics like PSNR, SSIM and quality of de-noised image is significantly improved in the proposed implementation. As hard thresholding algorithm eliminates dissimilar patches before PNLM de-noising the elimination of dissimilar patches more for neighborhoods with more details and less for flat neighborhoods.  Further de-noising scheme can be extended to any of adaptive weighting process which may yield far better results and improved performance than proposed implementation.

References

[1] C. Tomasi and R. Mantachie, “Bilateral filtering for gray and   color   images,”   in         Computer   Vision,   1998.Sixth International Conference on. IEEE, 1998, pp.             839–846.

[2] A. Buades, B. Coll, and J.-M.Morel, “A review of image denoising algorithms, with       a new one,” Multiscale Modeling & Simulation, vol. 4, no. 2, pp. 490–530, 2005.

[3] “A non-local algorithm for image denoising,” in Computer Vision  and  Pattern             Recognition, 2005.  CVPR  2005.IEEE Computer Society Conference on, vol. 2.         IEEE, 2005, pp.60–65.

[4] Y. Wu, B. Tracey, P. Natarajan, and J. P. Noonan, “Probabilistic non-local means,”       Signal Processing Letters, IEEE, vol. 20, no. 8, pp. 763–766, 2013.

[5] R. Vignesh, B. T. Oh, and C.-C.Kuo, “Fast non-local means (nlm) computation             with probabilistic early termination,” Signal Processing Letters, IEEE, vol. 17, no.         3, pp. 277–280, 2010.

 

681 total views, no views today

Download PDF File

About the author: admin