It is challenging to visualize 3D ultrasound data and offer high-fidelity
anatomical information in a real-time manner. The outcomes always
have high speckle noise and low signal-to-noise ratio. Besides, the
complicated anatomical relationship among tissues also poses a
problem for visualization. Existing visualization systems are inadequate
for precise diagnosis.
We have thus developed a novel ultrasound visualization system,
leveraging a set of advanced image denoising algorithms and volume
rendering techniques. Offering enhanced anatomical details and depth
information, the system can be used as a real-time 3D fetal ultrasound
visualization system. It includes our new 3D denoising algorithm to
effectively remove the speckle noise while preserving anatomical
features in the ultrasound data. Secondly, our novel automated
transfer function (TF) design scheme based on affinity propagation (AP)
clustering algorithm can automatically differentiate different tissues, so
that appropriate optical properties for visualization can be assigned.
Thirdly, our perception enhanced global illumination model based on
volumetric photon mapping can further enhance the details.
The proposed system can greatly increase the competitiveness of
ultrasound devices, and hence has great commercial potential.