Image Denoising in the Wavelet Domain Using Wiener
- Slides: 8
Image Denoising in the Wavelet Domain Using Wiener Filtering Nevine Jacob – Aline Martin ECE 533 Project – Fall 2004 nmjacob@wisc. edu – alinemartin@wisc. edu 12/14/2004 Image Denoising in the Wavelet Domain using Wiener Filtering
Problem statement Y = X + W = Y: Noisy image + W: White Gaussian noise X: Original image Assumptions 1/ X is unknown 2/ X and W are uncorrelated Goal: recover X from Y 3/ noise variance may be unknown 12/14/2004 Image Denoising in the Wavelet Domain using Wiener Filtering 2
Wiener Filter in the Wavelet domain 3 steps: 1/ wavelet transform 2/ Wiener Filter on the wavelet coefficients 3/ Inverse wavelet transform WF Noisy Im 12/14/2004 Level 1 – Wiener Filtered coefficients Image Denoising in the Wavelet Domain using Wiener Filtering Denoised Im 3
Wiener Filter in the Wavelet domain Level 1 – Wiener Filtered coefficients : variance of the noise 12/14/2004 Image Denoising in the Wavelet Domain using Wiener Filtering 4
Simulation Results • Wavelet domain: WF vs Thresholding Wiener Filter wavelet domain MSE = 110 12/14/2004 Soft Thresholding MSE = 140 Image Denoising in the Wavelet Domain using Wiener Filtering Hard Thresholding MSE = 175 5
Simulation Results • WF: wavelet domain vs Fourier Domain Wiener Filter wavelet domain MSE = 110 12/14/2004 Global Wiener Filter MSE = 115 Image Denoising in the Wavelet Domain using Wiener Filtering Local Wiener Filter MSE = 75 6
Simulation Results • MSE 12/14/2004 Image Denoising in the Wavelet Domain using Wiener Filtering 7
Conclusion Wiener Filter in the Wavelet domain performs better than thresholding methods and Wiener Filter in the Fourier Domain Improve denoising along the edges Need for a better quantitative criteria 12/14/2004 Image Denoising in the Wavelet Domain using Wiener Filtering 8
