Abstract: A multi-noise model-based denoising method for radiographic image is proposed, in which wavelet transform and median filtering are used. The composition of the image noise is analyzed, and a multi-noise model is established. The variance of noise is calculated in terms of the multi-noise model, and then the BayesShrink threshold is calculated , which solves the problem of that the Donoho’s noise algorithm is invalidated since the wavelet coefficients do not obey the generalized Gaussian distribution. A median filtering is used to refine the result obtained by the wavelet transform to eliminate the image distortion caused by the BayesShrink thresholding. Radiographic images are used to verify the effectiveness of the proposed method. Experiments show that the performance of the proposed method is better than those of OracleShrink and SureShrink.

Key words: information processing technology, multi-noise model, wavelet, image denoising