Abstract
A new computer-based approach to processing images, called Super Resolution, accomplishes the seemingly impossible—increasing a digital image's resolution so that elements which cannot otherwise be seen become visible. Such topics, appearing to border on magic but based on sound mathematical and physical principles, were discussed at a 23 February 1999 meeting organized by Malur Sundareshan at the University of Arizona. Originally developed as tools for such tasks as restoring blurred images from the flawed lens of the Hubble Space Telescope, the new algorithms are being used for everything from detecting previously unseen oil slicks on the surface of the ocean to extracting information from blurry old photographs. Image enhancement by Super Resolution could conceivably enable researchers to track bullet paths on frames of the famous Zapruder film of the Kennedy assassination. To understand how Super Resolution works, consider a digital image (for example, computerized scan of a photograph). It is composed of a tiny grid of dots called pixels that, when put together, recreate the image. Simple resolution-increasing techniques (such as duplicating neighboring pixels) do not, however, add more information to the image. Super Resolution, on the other hand, attempts to reconstruct the original scene that gave rise to the image. In doing so, the technique effectively rebuilds the picture at a new, higher resolution than the parent image. As an example, imagine one blurred photo showing two faces and a second sharply focused photo showing one of the same faces. Super Resolution algorithms would identify the mathematical transformations necessary to convert the blurred face in the first photo to the sharp image of the same face in the second photo. These operations, when applied to the entire blurred photo, define a means of reconstructing the second face as well. Thus, researchers who know the detailed optical properties of a telescope and the basic composition of the objects in its field of view can apply the technique to visualize objects beyond the telescope's resolving power. This is merely one simple application of Super Resolution. Researchers have combined numerous sophisticated mathematical operations to give the technique tremendously powerful restoration and resolution-enhancing capabilities. Knowledge about the visual properties of the objects in the image is essential for the process to work. The limitations of the original imaging system must also be taken into account. Logical constraints, such as limiting an object's apparent brightness, are employed in building the algorithm. Further considerations may include an object's size, color, motion, composition, reflectance, or texture. The more detailed these data and assumptions are, the better the image can be reconstructed. The processing algorithms employ Fourier transformation, noise filtering, and other mathematical tools. The product of their analysis is a reconstructed scene with the maximum likelihood of having been the source for the original image. Super Resolution is applicable to all types of spectral data, not just visible light. One area of current interest is for frequencies from 30 to 120 GHz, generally called the millimeter-wave band of the spectrum. Imaging resolution from this range of frequencies is low, but has the advantage of penetrating clouds, smoke, clothing (showing concealed weapons), or even buildings. Emissions around 95 GHz have tremendous potential for acting like a sophisticated kind of “x-ray vision” because of their ability to penetrate coverings that block the passage of visible light. There are two ways to image an object in this frequency range. The first method relies on beaming 95 GHz frequencies at an object and analyzing the results with a detector. The more desirable, passive imaging technique collects 95 GHz emissions from objects using a device called a radiometer and provides image data without the need for externally applied radiation. One limitation to using 95 GHz radiation in conjunction with Super Resolution for practical purposes, such as airport security, is the relatively slow time (approximately one minute) for acquiring and processing images. Current work is focused on using Super Resolution to improve the quality of these images and to reduce the computing time required for processing them.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.