LMMSE-Based Image Denoising in Nonsubsampled Contourlet Transform Domain
This paper proposes a nonsubsampled contourlet transform (NSCT) based multiscale linear minimum mean square-error estimation (LMMSE) scheme for image denoising. The contourlet transform is a new extension of the wavelet transform that provides a multi-resolution and multi-direction analysis for two dimension images. The NSCT expansion is composed of basis images oriented at various directions in multiple scales, with flexible aspect ratios. Given this rich set of basis images, the NSCT transform effectively captures smooth contours that are the dominant feature in natural images. To investigate the strong interscale dependencies of NSCT, we combine pixels at the same spatial location across scales as a vector and apply LMMSE to the vector. Experimental results show that the proposed approach outperforms wavelet method and contourlet based method both visually and in terns of the peak signal to noise ratio (PSNR) values at most cases.
KeywordsDenoising NSCT LMMSE PSNR
- 6.Lang, M., Guo, H., Odegard, J.E.: Noise reduction Using Undecimated Discrete wavelet transform. IEEE Signal Processing Letters (1995)Google Scholar
- 7.Coifman, R.R., Donoho, L.D.: Translation Invariant De-noising. In: Antoniadis, A., Oppenheimm, G. (eds.) Wavelet and Statistics, pp. 125–150. Springer, Berlin (1995)Google Scholar
- 18.Li, X., Orchard, M.: Spatially Adaptive Image Denoising Under Overcomplete Expansion. In: Int. Conf. Image Process., Vancouver, Canada, pp. 300–303 (2000)Google Scholar