Presented at the 2000 APIII Conference Return to 2000 Abstract Index
MANAGING DIGITAL IMAGE LIBRARIES WITH IMAGESHARK™ IMAGE SERVER
University of
Alabama at Birmingham
Birmingham, Alabama
Peter G. Anderson, DVM, PhD
Kristopher N. Jones, Dwain E. Woode, Kristina T.C. Panizzi,
Peter G. Anderson
Department of Pathology, University of Alabama at Birmingham
Background: The management of large online digital image libraries is complicated by the need to provide multiple versions of the same image for differing purposes—ranging from thumbnails to large presentation-quality images. This problem is compounded when the original source images are themselves of different sizes, resolutions, and formats. The Pathology Education Instructional Resource (PEIR) Digital Library contains over 50,000 images representing an integration of several existent image databases as well as new images being actively contributed via a multi-institutional upload feature. Attempts at manually standardizing source image size and format have been only partially successful and are expensive in terms of time, personnel costs, and disc storage space. The purpose of this project was to evaluate ACD System’s ImageShark™ Web Image Server, which has been designed to reduce digital library management costs by its ability to generate images of varying size, resolution, and format from a single unmodified source file.
Design: ImageShark™ is an ISAPI extension DLL designed to handle CGI requests on a Microsoft IIS Web server. HTTP requests containing query string variables denoting the target image’s size, resolution, compression ratio, and format are passed by the Web server to the DLL which then generates the requested image and returns it to the Web server. In addition, image enhancements such as sharpening, leveling, and watermarking may also be specified. Any image capable of being served by the Web server can be manipulated by the DLL—thus images may reside in disparate directories.
Results: By allowing only a single source image to be maintained on the server, image preparation time, personnel cost, and amount of required storage space were all reduced. In addition, ImageShark™ automatically compensated for differences in source image size more consistently than our previous manual efforts. Having the output images generated on the fly was not without performance costs, however—the time for the server to return thumbnails increased from an average of 33 to 506 milliseconds while the return time for the highest-quality images increased from 388 to 1819 milliseconds.
Conclusion: ImageShark™ reduced the costs of managing our large digital library and improved the consistency of image delivery while allowing us to use disparate image source files from various contributors. Though it takes longer to generate output images on the fly, these delays are mostly transparent to the user; with twenty thumbnails on a search result page, user scrolling masks the cascading download of images. Further, the program temporarily caches generated images (length of time is configurable) to increase speed of return if called again.
