Abstract:

To overcome the defect of image denoising based on Donoho’s threshold of wavelet domain, a self-adaptive method for image denoising based on stationary wavelet transformation was proposed. According to different subband characteristics, a new scale parameter equation was defined todetermine the optimal thresholds adapting to each step scale, and the thresholds of the detail signals of each level generated by stationary wavelet transformation were obtained.The method not only well depresses the Gibbs impact well, but also supplies a gap of orthogonal wavelet transformation so as to retain the MR image details when wiping offnoise. Experimental results indicate that the proposed method is more effective than discrete orthogonal wavelet transformation with respect to performance and visual effect.

Key words: orthogonal wavelet transform; self-adaptive threshold; MR image denoising; stationary wavelet transform