Despeckling Medical Ultrasound Images Based on Blind Deconvolution and Maximum-Likelihood Estimation

In view of the inherent speckle noise in medical ultrasound images, a novel method is proposed based on the blind deconvolution and the Maximum-Likelihood (ML) filtering. Firstly, the point spread function (PSF) is blindly estimated based on the complex-cepstrum filtering; then the PSF is input into a Wiener Filter in order to recover the tissue reflectivity. Secondly, the reflectivity is reasonably assumed to be corrupted by the multiplicative noise obeying the Rayleigh distribution, and an adaptive ML estimation approach is utilized to eliminate such noise. Finally, the noise-free tissue echogenicity map is obtained. Simulation and in vivo results indicate that the proposed method strikes an ideal balance between the edge preservation and the noise suppression, and its quantitative assessment outperforms existed despeckling algorithms.