Spatially adaptive multiscale thresholding for speckle and mixed noise removal
Hengameh Keshavarz, M. Edward Jernigan, Javad Ahmadi‐Shokouh · 2005
Summary form only given. This paper presents an adaptive thresholding technique in the wavelet-transform domain to remove multiplicative noise. Unlike other speckle reduction methods, this approach requires no a priori knowledge of the noise distribution. Hence, this proposed approach is applicable also for non-speckle noise, such as mixed noise. The proposed algorithm: (1) applies the wavelet transform on noisy images; (2) computes the wavelet coefficients' variances for detail sub-images; (3) identifies noisy wavelet coefficients via the analysis of variance (ANOVA) method; (4) denoises approximation coefficients via low pass filtering; and (5) reconstructs the denoised images via the inverse wavelet transform. Simulations verify this technique's efficacy in speckle and mixed-noise removal and demonstrates this technique's superiority over some other adaptive schemes.