Image processing workstation software development and feature size measurement methods for NDE X-ray images

Abstract

State-of-the-art image scanning devices with spatial resolution down to 25 microns and 8-bit intensity resolution currently are sold for $40,000 to $60,000; systems with 12-bit resolution for more demanding applications are priced yet higher. Also, increased image resolution means larger volumes of data to process, which leads to increased speed requirements (and thus increased cost) for the processing system to keep overall inspection times reasonable. Consequently, demanding applications still tend to be served by special-purpose expensive hardware. One NDE radiograph image processing system receiving much attention at this time is the Scan IV system by DuPont. This system consists of a high-resolution (35 microns spatial resolution and 3.5 decades light intensity dynamic range) digitizer, a workstation-class computer augmented by several add-on image processing boards, high-capacity (~2 gigabyte), high-speed optical disk drives, a video signal digitizer for incorporating real-time video images into the system, 3 image display CRTs, and a high-resolution digital film recorder for film hardcopy output (Eizember, 1990). This system is presently sold for hundreds of thousands of dollars and reportedly requires several person-months of time to set up and get running. Other radiographic image processing systems have been developed at Ohio State University, the Army Materials Technology Laboratory, and the Electric Power Research Institute, among others (Sheppard, 1987). The high price of such systems keeps their sales volume low, and so highperformance radiograph image processing is presently not a commodity. Development of software products for these systems tends to proceed slowly, with custom work being done for each customer and with software not being portable between different high-end systems. There seems to be widespread agreement that the "traditional approach of using custom hardware and software to address the imaging applications has actually retarded the growth of new imaging technology by keeping prices high and not addressing the issue of standards conformity required to spur application development" (Pfeiffer, 1990, p. 36). The imaging industry has begun to respond to the difficulties presented by high-priced custom image processing systems. Pfeiffer (1990) argues that

Automated image inspection using wavelet decomposition and fuzzy rule-based classifier

“…Therefore, it is necessary to built a graphic user interface to fulfill the interactive training, image processing and display. The image inspection system software and its graphic user interface are developed under a Unix and X window environment named HAPXM, which is a new version of the package named HAPPI [Brown, 1992].…”

Automated flaw detection method for X-ray images in nondestructive evaluation

“…5 shows the noise filtering results for the crack type flaw. As before, the top row shows the results of the adaptive smoothing filter using window sizes of 3, 5, and 7, respectively.…”

“…(3)Current topological descriptor (4). Number of pixels added last time (5). Number of pixels added time before that (6).…”