Author : Lina Desale 1
Date of Publication :21st August 2017
Abstract: Utmost prevailing practical digital video denoising techniques depend upon a traditional statistical model of image noise, such as Independent and identically distributed random variables- Gaussian noise, which is often violated in real life scenarios. For example, following foremost sources of video noise with dissimilar statistical distributions have been identified: photon shot noise, fixed pattern noise, amplifier noise, dark current noise and quantization noise. Performance of prevailing video denoising algorithms will severely degrade when the algorithm is applied on those images with noises evident from multiple sources. In this paper, a dictionary learning based scheme is proposed which computes the basis function adaptively from the first input image frames per fifty frames. Unlike other classical approaches like wavelet or contourlet transforms where the mother wavelet/basis functions are constant. If the mother wavelet/basis function is constant it is more likely that it will fail to capture the minuscule noise details from real life images. Therefore, the basis function is learnt from first frame itself. The dictionary learning method provides sparse representation of the image. Here, hard thresholding algorithm is applied to compute the denoised frame.
Reference :
-
[1] Buades, A., Coll, B. and Morel, J.M., 2005, June. A non-local algorithm for image denoising. In Computer Vision and Pattern Recognition, 2005. CVPR 2005. IEEE Computer Society Conference on (Vol. 2, pp. 60-65). IEEE.
[2] A. Buades, J. L. Lisani, and M. Miladinović, "Patch-Based Video Denoising With Optical Flow Estimation," IEEE Transactions on Image Processing, vol. 25, pp. 2573-2586, 2016.
[3] G. Chen, J. Zhang, D. Li, and H. Chen , "The Robust Kronwecker product video denoising based on fractional-order total variation model," Signal Processing, vol. 119, pp. 1-20, 2016.
[4] R. Almahdi and R. C. Hardie, "Recursive non-local means filter for video denoising with Poisson-Gaussian noise," in 2016 IEEE National Aerospace and Electronics Conference (NAECON) and Ohio Innovation Summit (OIS), 2016, pp. 318- 322.
[5] R. A. Ali and R. C. Hardie, "Recursive nonlocal means filter for video denoising," EURASIP Journal on Image and Video Processing, vol. 2017, p. 29, 2017.
[6] M. Khalid, P. S. Sethu, and R. Sethunadh, "Optimised surfacelet transform based approach for video denoising," in 2016 International Conference on Inventive Computation Technologies (ICICT), 2016, pp. 1-5
[7] H. Guo and N. Vaswani, "Video denoising via online sparse and low-rank matrix decomposition," in 2016 IEEE Statistical Signal Processing Workshop (SSP), 2016, pp. 1-5.
[8] J.-S. Lee, “Digital image smoothing and the sigma filter,” Computer Vision, Graphics & Image Processing, vol. 24, no. 2, pp. 255–269, 1983.
[9] D. Zhang, J.-W. Han, O.-J. Kwon, H.-M. Nam, and S.-J. Ko, “A saliency based noise reduction method for digital TV,” in Proceedings of the IEEE International Conference on Consumer Electronics (ICCE ’11), pp. 743–744, Las Vegas, Nev, USA, January 2011.
[10] A. A. Yahya, J. Tan, and L. Li, “An amalgam method based on anisotropic diffusion and temporal filtering for video denoising,” Journal of Computational Information Systems, vol. 11, no. 17, pp. 6467–6475, 2015.
[11] Ali Abdullah Yahya, Jieqing Tan, and Lian Li, “Video Noise Reduction Method Using Adaptive Spatial-Temporal Filtering,” Discrete Dynamics in Nature and Society, vol. 2015, Article ID 351763, 10 pages, 2015. doi:10.1155/2015/351763