Blind deconvolution of ultrasound images

We address the problem of improving the resolution of ultrasound images using blind deconvolution. The transducer measurement, that forms the ultrasound image, can be expressed as the convolution of two terms, the tissue response and the ultrasonic system response, plus additive noise. The quantity of interest is the tissue response, however, due to the convolution operation, we measure a blurred version of it, which obscures the fine details in the image. Our goal is to remove the blurring caused by the ultrasonic system, in order to enhance the diagnostic quality of the ultrasound image. Under the assumption that speckle behaves as an i.i.d. zero-mean non-Gaussian process, we were able to reconstruct the blurring kernel using bicepstrum operations on corresponding A-scan lines. Based on the estimated blurring kernel we performed deconvolution on the measured image. Preliminary results obtained from ultrasound images of a tissue mimicking phantom indicate significant resolution improvement.