The return of the DSP
n many low-end image-processing applications, fast Pentium-based PCs or graphics workstations are providing all the power required to process images. But for more demanding applications, the digital signal processor (DSP) is coming to the forefront as a mid-range solution. Off-the-shelf boards are providing significant power in many industrial, medical, and military imaging tasks.
One medical application that uses programmable DSPs is ultrasound. As contributing editor Larry Brown explains, the modality has long been used as a noninvasive way of imaging the human body (p. 28). Now, by using a sophisticated PC-based add-in processor and custom software, a team at the University of Washington has helped Siemens develop a DSP-based ultrasound system.
As the marine equivalent of aerial photography, sidescan sonar produces images that are pieced together to produce a mosaic of the ocean floor. Here, too, DSP boards are finding inroads. Contributing editor John Haystead explains that researchers on the USGS Atlantic Geology Team in Woods Hole, MA, are using a number of sonar transducer systems to image the Exclusive Economic Zone, a 200-mile region of the US coastline (p. 14). Researchers are using PC-based systems, add-in DSP boards, and off-the-shelf software.
DSPs are also finding their way into image-processing systems at the crime laboratory of the Metro-Dade, FL, Police Department. John Haystead reports that the laboratory is using a host of image-processing tools and technologies, and forensic scientists are analyzing everything from narcotics, DNA samples, gunshot residue, and bullet casings to glass fragments and paint residue (p. 22).
While many vision systems still rely on standard programming practices, others are using more novel methods of understanding images. One such method, fuzzy logic, can be used to construct modular and reusable imaging code for image interpretation and classification. Contributing editor Barry Phillips explains that multiple units of knowledge, represented by fuzzy logic functions, can be combined to make classification decisions, closely mimicking the human decision-making process (p. 34).
But to develop a novel vision system requires more than processing power and software. Lenses also play an important role. In his article beginning on p. 42, Randal Chinnock explains the intricacies of choosing the right optical design software. Not only does he examine some of the more popular off-the-shelf packages, he also analyzes which software may be best for your specialized optical design.
Andy Wilson, Editor