University of Minnesota Driven to Discover
U of MNUniversity of Minnesota
Center for Transportation Studies

Programs & Labs

ITS Sensor Header

Fall 2000

Zooming in on better electronic images

Photo of Vladimir Cherkassky

Professor Vladimir Cherkassky

Digital imaging is one of the most important techniques for distributing information about the transportation system. The use of high-resolution digital images is hampered, however, by the low speed of typical Internet connections. Even with the gradual arrival of high-speed, high-bandwidth connections, delivering large and detailed photographic images over the Internet continues to be a challenge. But current research by Vladimir Cherkassky, professor of electrical engineering at the University of Minnesota, may improve the situation.

DOT's Office of Land Management (OLM) is funding the project, Image Compression for Storage and Transmission of Digitized Images, with the goal of developing a superior method for the storage and delivery of large images in electronic form. Specifically, the OLM hopes to use techniques developed by Cherkassky's team to make aerial photographs and other large, complex images available to users over the Internet. An additional benefit would be smaller electronic files that require less storage space.

The need for image compression

Although the words you are reading can be stored on a computer as a relatively small text file, representing a complex visual image such as a photograph in electronic form requires processing and storing a huge amount of data. As the maximum display size of the image is increased, the amount of data required increases.

For this reason, most digital image formats rely on some type of compression to keep image files small enough for convenient use. Images are compressed by specialized mathematical operations; the ultimate goal is to reproduce as much detail as possible while reducing the total amount of information that must be encoded in the image file.

Cherkassky's research focuses on improving the mathematical methods computers use to compress images. His group compared two methods of image compression: Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT). The JPEG and MPEG image formats popular on the World Wide Web implement the "traditional" DCT method; the DWT method is less widely used.

"Accessing large image files over the Web requires image compression capability, zooming capability, and incremental transmission capability," Cherkassky says. "Our research addresses all these issues and their practical software implementation."

While both methods offer similar performance in terms of compression ratio, the DWT or "wavelet" method attracted the researchers because it could be used to selectively decompress parts of the image selected by the viewer—in effect, allowing viewers to "zoom in" to examine small details, or zoom back out to view large areas.

This zooming functionality is particularly important to Mn/DOT, because it will greatly enhance a remote userís ability to navigate and search aerial photographs. Using traditional compression methods, the entire image would have to be decompressed to view it at maximum resolution.

The first phase of the research project involved adapting a commercially available software package that uses DWT compression to meet the specific needs of the OLM. A prototype system is currently operational, providing a limited number of sample aerial images via a Web site. Users can click on an image to zoom in through several steps of magnification, without seeing significant loss of detail.

For phase two, Cherkassky and his team are focused on developing their own DWT algorithm, incorporating another important feature: incremental compression, or the ability to transmit a low-resolution version of an image and follow up with progressively more detail. This capability, not offered by the commercial software used in phase one, will be especially important for the rapid distribution of large images over the Internet.

The new compression algorithm Cherkassky's team is developing must fulfill a difficult mission: reduce electronic file size while maintaining image quality, and also allow for progressive transmission and incremental compression of large images.

To accomplish this, Cherkassky is building his new algorithm around an approach called wavelet thresholding, which has been widely studied by researchers and shown to outperform traditional linear approaches to image compression. The term 'thresholding' refers to the method of discarding signal information below a specified threshold, thereby reducing electronic ìnoiseî which does not add to image quality but takes up valuable space in the electronic file. The new compression algorithm will be a unique implementation of wavelet thresholding, with innovative approaches to encoding the signal and setting the threshold level.

In a world increasingly reliant on electronic media for the exchange of information, the work of Cherkassky's team has far-reaching implications, since transmission and compression of large image files is a common task in many applications. Ultimately, much of the visual material transmitted over the Internet will benefit from better compression schemes.