Blind deconvolution of ultrasound images using partial spectral information and sparsity constraints

The ability to enhance the resolution of ultrasound images by deconvolution relies mainly on knowledge of the system's point spread function (PSF). The PSF is usually unavailable due to its spatial nonstationarity, and it needs to be estimated from the acquired data. Cepstrum-based techniques are unable to correctly restore the spectral phase of the PSF. On the other hand, some iterative (optimization) methods, which simultaneously estimate the PSF and the reflectivity distribution of the tissue, converge slowly, due to insufficient prior information. In this study, it is shown that the performance of the optimization-methods can be considerably improved, by imposing additional constraints, which represent some partial information about the PSF spectrum, e.g. its power spectrum. The method is tested on simulated data of soft tissue, which includes randomly distributed strong specular reflectors. The SNR is improved by approximately 18 dB, the axial resolution by 6 dB and the lateral resolution by 11 dB. These results justify further in-vitro studies.