Denoising algorithm with multiple thresholds based on the relativity exponents of wavelet coefficients

Firstly the wavelet coefficients are estimated by NeighShrink scheme and then 0-1 binary character coefficients are obtained. Based on the above binary coefficients, the intra and inter relativity exponents are defined respectively for each wavelet coefficients. Finally decision coefficients are determined according to the magnitudes of the intra and inter relativity exponents. This algorithm overcomes the over-reserved or over-shrinked disadvantages of simple threshold methods in that the decision coefficients are obtained by multiple criteria. Besides, large impulse noises can be removed by the presented algorithm because it takes signal features into consideration. The experimental results show that the proposed algorithm is computationally efficient and practical.