Low-complexity image denoising based on statistical modeling ofwavelet coefficients

摘要

We introduce a simple spatially adaptive statistical model fornwavelet image coefficients and apply it to image denoising. Our model isninspired by a recent wavelet image compression algorithm, thenestimation-quantization (EQ) coder. We model wavelet image coefficientsnas zero-mean Gaussian random variables with high local correlation. Wenassume a marginal prior distribution on wavelet coefficients variancesnand estimate them using an approximate maximum a posteriori probabilitynrule. Then we apply an approximate minimum mean squared error estimationnprocedure to restore the noisy wavelet image coefficients. Despite thensimplicity of our method, both in its concept and implementation, ourndenoising results are among the best reported in the literature