Abstract: As the main prevailing denoising method, how the threshold function works and what the threshold value is are of the greatest importance. According to the interscale and intrascale dependencies of the coefficients in the non-subsampled Contourlet transform domain, and considering the change of coefficient’s aggregation with different directional subbands in the same scale, a novel non-subsampled Contourlet transform denoising scheme using the directional property (AD_NSCT) is proposed. This scheme can lead to enhanced estimation results for images that are corrupted with additive Gaussian noise over a wide range of noise variance. To evaluate the performance of the proposed algorithms, simulation results are compared with those by the algorithms, such as wavelet threshold, Contourlet transform threshold and non-subsampled Contourlet transform threshold for image denoising. The simulation results indicate that the proposed method outperforms the others 0.5~3.3dB in the PSNR, and keep a better visual result in edges information reservation as well.

Key words: threshhold function, wavelet transform, non-subsampled Contourlet transform, scale dependency, denoising