Abstract: In this paper, we formulate a variational framework for the deconvolution of ultrasound images obtained from pulsed-echo linear array transducers. These images, obtained as a result of acoustic response of soft-biological tissues are called Tissue Reflectivity Function (TRF). TRF suffers from speckle patterns due to non-homogeneous nature of soft tissues being interrogated. The interference of reflected beam develops random patches of bright and dark spots. The reduced spatial resolution in the axial direction worsened as a function of distance from the transducer probe affects the diagnostic significance of ultrasonography. We design an optimization framework with consideration to Gaussian and speckle noise characteristics in the form of two data fidelity terms. To preserve edges during the iterative reconstruction process, we introduce Total Variation (TV) regularization term as well. We have conducted experiments on artificially corrupted synthetic data and simulated and real ultrasound data. Experimental results using several metrics supported by the visual results show improvement over state-of-the-art techniques for ultrasound image restoration. (Online) 

Index Terms— Ultrasound imaging, deconvolution, Gaussian-Speckle Noise, Tissue Reflectivity Function.

Cite this as: Aetesam, Hazique, and Suman K. Maji. "Ultrasound image deconvolution adapted to gaussian and speckle noise statistics." 2020 7th International Conference on Signal Processing and Integrated Networks (SPIN). IEEE, 2020.