Fast processors speed image visualization and analysis

Because of the effort required to program them, parallel processors have never taken hold in general-purpose computing. But in embedded systems, parallel processors are providing the power to develop novel applications.

Jan 1st, 1997

Fast processors speed image visualization and analysis

Because of the effort required to program them, parallel processors have never taken hold in general-purpose computing. But in embedded systems, parallel processors are providing the power to develop novel applications.

At the University of Chicago and Argonne National Laboratory, a prototype 3-D optical microscopy system has been developed that overcomes the computational demands of 3-D acquisition, reconstruction, and display. Contributing editor Rick Nelson explains that by using parallel computing and high-speed networks, the system can visualize 3-D cell images in a virtual-reality environment (p. 30).

Medical applications are also benefiting from parallel processing. Magnetic-resonance-imaging (MRI) technology requires high-performance image-processing hardware and software that can maximize the resolution and manipulation of medical images in real time. Contributing editor John Haystead examines how MRI system designers and producers have addressed these concerns with off-the-shelf image processing boards, commercial bus architectures, and data links (p. 38).

Designers of medical imaging systems should also read the Spotlight on Advanced Technology (p. 42). There, contributing editor Barry Phillips describes how pattern recognition is being used to analyze MRI, digitized x-ray, and pathology data and discusses various classifiers including artificial neural networks and fuzzy-based methods.

With the FAA pressing hard for new explosive-detection capabilities for baggage inspection, airport security systems are also emerging as an application for parallel processing. On p. 24, John Haystead describes a system that combines CT and x-ray imaging using an off-the shelf image processor.

But parallel processing does has its rivals. Typical web-inspection systems use stationary line-scan cameras and the motion of material being inspected to form 2-D images for later analysis by computers. In these types of applications, pipelined image processing may be a better choice. Eric Peterson of Proteus Applied Technologies (San Francisco, CA) describes a web-inspection system using off-the shelf hardware and software to find small defects in large areas of coated glass (p. 18).

In machine-vision applications, lighting also plays a crucial role. In his article on p. 50, Dave Wilson looks at different types of illumination equipment available and shows systems that make best use of each technology.

Andy Wilson Editor

andyw@pennwell.com

More in Home