Image denoising using learned overcomplete representations

We describe a method for learning sparse multiscale image representations using a sparse prior distribution over the basis function coefficients. The prior consists of a mixture of a Gaussian and a Dirac delta function, and thus encourages coefficients to have exact zero values. Coefficients for an image are computed by sampling from the resulting posterior distribution with a Gibbs sampler. Denoising using the learned image model is demonstrated for some standard test images, with results that compare favorably with other denoising methods.

[1]  Edward H. Adelson,et al.  The Design and Use of Steerable Filters , 1991, IEEE Trans. Pattern Anal. Mach. Intell..

[2]  Edward H. Adelson,et al.  Shiftable multiscale transforms , 1992, IEEE Trans. Inf. Theory.

[3]  D. Donoho,et al.  Translation-Invariant De-Noising , 1995 .

[4]  William T. Freeman,et al.  Presented at: 2nd Annual IEEE International Conference on Image , 1995 .

[5]  Edward H. Adelson,et al.  Noise removal via Bayesian wavelet coring , 1996, Proceedings of 3rd IEEE International Conference on Image Processing.

[6]  H. Chipman,et al.  Adaptive Bayesian Wavelet Shrinkage , 1997 .

[7]  B. Silverman,et al.  Wavelet thresholding via a Bayesian approach , 1998 .

[8]  Robert D. Nowak,et al.  Wavelet-based statistical signal processing using hidden Markov models , 1998, IEEE Trans. Signal Process..

[9]  Martin Vetterli,et al.  Spatially adaptive wavelet thresholding with context modeling for image denoising , 2000, IEEE Trans. Image Process..