Real-time echocardiographic noise reduction, border extraction, and velocity derivation

The authors describe initial experience in echocardiographic image processing utilizing PIPE, a high-speed eight-stage array processor capable of performing multiple sequential algorithms at the real-time video field rate. The algorithms implemented include noise reduction with recursive temporal filtering, Gaussian spatial filtering, and edge-preserving smoothing; edge enhancement with the Sobel for magnitude and angle operators and difference-of-Gaussian convolutions; and velocity quantification with convected activation (a method for deriving normal motion from a sequence of binary edge maps described by A. M. Waxman et al., Proc. IEEE Conf. Comput. Vision Pattern Recognition, June 1988). It is shown that PIPE is a promising low-level engine for echocardiographic image processing. It remains, however, to be fully integrated with a high-level computer.