Model based estimation of speckle filter for improved suppression of multiple scattering noise with dual band ultrasound pulse complexes
Abstract
In ultrasound imaging, multiple scattering noise can be severely damaging to the image quality. SURF imaging is a non-linear imaging method that can be used to suppress the multiple scattering noise, and thus improve the image quality. The suppression is done by processing methods involving a delay correction and a speckle correction. In this thesis, methods for speckle correction have been investigated on simulated signals.
The underlying effects that cause the speckle change in SURF signals have been investigated, with emphasis on an effect known as pulse form distortion. A model for describing the pulse form distortion have been tested, and proved to perform well, with the error between the model and the simulations typically well under -30dB. It was also shown a linear relationship between two of the parameters in the model, which can possibly be utilized for finding the phase of the pulse form distortion.
Three different methods for performing speckle correction have been studied, and compared to traditional methods that only uses a delay correction. Two of them yielded positive results, the average non-linear phase method and the physical speckle model. The physical speckle model is very complex however, and the performance dropped much when non-ideal model parameters were used. Therefore, the average non-linear phase method is concluded to be the most promising, although it must be tested in more realistic scenarios. The achieved improvement in signal-to-noise ratio with this method was shown to be around 3-5dB better than the traditional methods.