How do you think the new GigE standards will influence the machine vision industry?
Respond or ask your question now!
Awaiting the Reformation API Standardization Has Promised
Digital imaging has long promised to reform a wide range of applications, but this has yet to occur. Now, with API Standardization, open interfacing and true interoperability may yet be possible...
by Lee J. Nelson
Imaging is in my blood! My professional career has developed alongside the commercial image processing industry. For the almost twenty-five years, I've been involved with the discipline, digital imaging has promised to revolutionize applications as diverse as remote sensing, biometrics, medicine and entertainment. Truth be told, however, most of us still are awaiting the dramatic reformation.
With the possible exception of diagnostic medical imaging, where replicate applications thrive, nearly every high-end solution proves to be either a made-to-order system or a unique amalgam of off-the-shelf components with its requisite "glue-code." Each consumer expects a solution perfectly suited to his/her needs. While vendors endeavor to accommodate patrons' desires, they unhappily ruminate about those projects, long devoid of profit and likely to require maintenance into perpetuity. They find it practically impossible to reuse and, therefore, res-ell the results of such customization. Even now, replete with dozens of imaging standards (all duly noted in the pages of this magazine), we must acknowledge the lack of standardization.
In medical imaging, applications are fairly routine and well-defined. Multi-modality images are acquired, processed, displayed, routed, annotated and archived. Sometimes hardcopy is produced. The set of sensors is finite and known, as are output devices and storage media. Operator interfaces accommodate structured environments, giving physicians and surgeons access to patient imagery, written reports and audio summaries, from radiologists or nuclear medicine specialists, via Picture Archival and Communication Systems (PACS).
More recent work in PACS design centers on image recovery and reconstruction with popular retrieval methods based on keywords, but keywords don't necessarily directly reflect image content; rather, they're descriptors used to characterize image properties. It is hard to institute a complete, accurate and unambiguous lexicon that describes all image attributes for everyone. Limitations of the traditional keyword approach are giving way to content-based image retrieval (CBIR). CBIR represents a more natural means of communication. Following a user-request, a CBIR-PACS searches for images possessing similar characteristics (texture, color, shape, density, etc.) and returns the relevant data.