Q+A CENTER

FEATURE

The Quest for a Practical Image Processing Standard

Use of image processing technology has increased dramatically in the last decade, bringing with it a growing interest in the establishment of a reference standard for image processing algorithms...

by Lisa Kemplerand Steve Eddins

May 2003

Besides the well-publicized applications in medical imaging and satellite reconnaissance, image processing is playing a central role in many other areas, including consumer electronics (digital cameras and displays), robotics and manufacturing (machine vision), environmental monitoring, security and access control. The perceived need for a global image processing standard became apparent as early as 1991, when experts within the International Organization for Standardization and the International Electrotechnical Commission formed a joint technical committee (JTC1 SC24) to develop such a standard. The committee included representatives from the United States (William K. Pratt, then with Sun Microsystems), The United Kingdom (Adrian F. Clark, University of Essex), and Germany (G. Rainer Hofmann, Fraunhofer Group).

The objective of this standards effort was to define a common architecture for imaging, as well as acceptable practices for data transfer and processing operations. An API for the planned standard called for a programmer's imaging kernel system (PIKS), to support most common operations on images. To simplify the API, data types were restricted. For example, image data had five dimensions-three spatial coordinates and one each for spectral components and time.

Leaving its door open for technology advances, the standard was, for the most part, completed in 1994 as ISO/IEC Standard No. 12087 and reapproved five years later. The third part of that three-part proposed standard-functional specification of image interchange-was published; however, a data-encoding document meant to accompany it was never completed. In its place, the computing and Internet worlds have successfully adopted several commercial image processing interchange formats, such as TIFF and JPEG.

In another attempt to develop an image processing standard, researchers in 1996 began developing VSIPL (Vector, Signal and Image Processing Library). VSIPL, developed by representatives from embedded signal and image processing hardware and software vendors, academia and government laboratories, resulted in an API and a mathematical library. The VSIPL library consists of matrix, linear algebra and signal processing functions. Similar to other standards, such as PIKS, it offers portability and object-based descriptions, among other characteristics.