The case for color is getting stronger all the time. And with prices dropping, the cost of getting started has never been lower.
Even if your bread-and-butter imaging applications use black-and-white documents, you probably daydream about color now and then. The stunning capabilities of today's technology make the allure of color imaging nearly irresistible.
Why just daydream? A compelling case can often be made for color in traditional black-and-white document imaging environments -- especially in companies with serious Web or corporate intranet ambitions.
The cost? It's dropping rapidly. Color scanners, printers and displays start at less than $500. While things are still pretty pricey at the high end, the cost of getting started is a bargain.
Go ahead. Add color to your world. Express yourself. Impress your customers. Color sells. Just make sure you compress your image files. Compression addresses the daunting demands that scanned color images place on two crucial components of your imaging system -- networks and storage.
1. Networks. The file size for an uncompressed scanned 8.5" x 11" color image can reach several megabytes. When you push several of these monsters across an Ethernet at the same time, your LAN performance degrades to a snail's pace.
Transmitting uncompressed color files over slower wide-area connections produces even bigger performance hits. That's why compression is almost mandatory for downloadable images stored on a Web site. If you make someone sit for several minutes waiting for an uncompressed color image to crawl down a dial-up connection, they won't visit your site again.
Compressed color images are 15-20 times smaller than their uncompressed versions -- between 50K and 100K, on average. That means faster transmission, less network overhead and more success in cyberspace.
2. Storage. Compressed images require fewer optical discs and jukeboxes for storage. It's easy to calculate the cost savings. If your images are a third of the original size when compressed, you need a third as much hardware for storage.
There are other less obvious storage advantages associated with image compression. Smaller files take less time to read from a storage device. Storing more files per disc means less disc swapping if your images are stored in a jukebox. Faster retrieval means less waiting when an image is requested. This lets your users be more productive.
Compression 101
The world of color compression is strewn with more jargon than almost any other area of imaging technology. And the mathematical frameworks for image compression schemes are mind-boggling.
Getting a handle on the fundamentals isn't difficult. A working knowledge is important, since compression is a key performance factor in any imaging system -- whether color or black-and-white. Knowing how to use compression technology effectively saves you money and improves productivity. Here's how it works.
Using complex algorithms, the compression process analyzes the bits that represent an image and then squeezes them into a smaller file. The amount of compression is represented by a ratio of the file's original size over its compressed size.
A compression ratio of 15:1 means the compressed image is 1/15 the original size. A 900 Kb file is reduced to about 60 Kb.
When the image is displayed or printed, the algorithm reverses the process and decompresses the image back to its original form.
Compression algorithms vary according to the type of image being compressed. With black-and-white (bitonal) images, the most common compression schemes are CCITT Group III and Group IV. Group III and Group IV are well-known because they're used in fax technology.
Group III bases its compression algorithm on reference lines within the image. Each line of the image is encoded based on how it differs from the most recent reference line. Since the number of bits required to store the changes in each line tend to be smaller than the total bits in the entire line, Group III eliminates redundant data. It stores only the information needed to reconstruct the original image. The primary difference between Group III and Group IV is that Group IV doesn't use reference lines. Most imaging systems use Group IV compression for bitonal images.
Both Group III and Group IV are "lossless" compression schemes. With lossless compression, every data bit in the image is accounted for to ensure the decompressed image is an exact duplicate of the original.
Lossless compression maintains the legibility of scanned documents. It also maintains the full integrity of the original document. This is important because companies want to avoid getting into document integrity disputes.
An alternative to Group III and Group IV is JBIG compression. This scheme, devised by the Joint Bi-Level Imaging Experts Group and approved by the International Organization for Standardization (ISO), provides better lossless compression ratios for bitonal images than Group III and IV. It also works much better with grayscale images.
Few applications support JBIG because Group III and IV were well-entrenched when the JBIG standard was approved in 1993. Recently, a coalition of companies representing a broad spectrum of the imaging industry was formed to bring JBIG out of the shadows and into the imaging mainstream (see sidebar).
Compression gets more complicated with color images. Unlike bitonal images, where bits are either black or white, color image bits contain information about color depth and brightness. Instead of one bit per pixel (as with black-and-white images) color images can have up to 24. The number of bits depends on the number of colors used in the image. Images that use 256 colors, use eight bits per pixel. Images that use 16 million colors (True Color images), use 24 bits.
Because they're so large, most color images are typically compressed with "lossy" compression schemes -- part of the image data is discarded to reduce the storage requirements of the original image file. Users decide how lossy they want the image to be by balancing file size with image quality.
With most color images, quality isn't affected very much when moderate amounts of data are tossed out. However, highly compressed color images typically include "artifacts," or image degradation produced by discarding data that's essential to accurately reproduce the original image.
Color Compression Technologies
Factors like compression ratio and image degradation depend on the type of image compression used. There are three major technologies available: JPEG, fractal and wavelet.
1. JPEG. The Joint Photographic Experts Group compression format is the most widely used tool for compressing high bit-level (24 bit) color images. With JPEG, you can compress a little or compress a lot -- it's adjustable.
If you need lossless compression, JPEG can provide it. But you won't reduce your file size overhead very much this way. Lossless JPEG typically produces low compression ratios in the range of 2:1 to 5:1.
Lossy versions of JPEG give you much better compression ratios. But you lose image quality if you try to stretch the ratios too far. 64:1 gives you very lean file sizes. You'll also probably produce a blocky "screen door" effect.
That's because JPEG compresses images by dividing an image into sections 8-pixels square. Using complex mathematics called discrete cosine transforms, the JPEG algorithm compares each section to the previous one and codes the difference. It also analyzes similarities between adjacent pixels within each square.
After this process is complete, odds and ends of image data are left over. This data is compressed using a lossless compression scheme called Huffman encoding.
To increase compression ratios, JPEG uses fewer pixels per square. Also, color data is sometimes compressed by averaging all the color values in each square. In highly-compressed JPEG images, these techniques often produce blocky artifacts.
Some JPEG compression vendors use proprietary techniques to minimize artifacts in over-compressed images. The best way to eliminate JPEG image degradation is to keep compression ratios at less than 30:1.
As color images become more common (thanks to the Web), industry experts say the compression marketplace will demand ratios higher than JPEG can provide without artifacts. There are contenders waiting in the wings. But they all lack one of JPEG's most important advantages: standards. With JPEG, a compressed image can be decompressed and viewed with any imaging software incorporating the ISO-approved JPEG specification.
That's one reason why JPEG is used by most high-resolution color applications on the World Wide Web. (GIF, a color compression scheme patented by Unisys, is used for the majority of low-resolution -- 8-bit and under -- color Web images.) Indeed, the Web is the driving force behind recent enhancements such as progressive JPEG, which was introduced last year to allow the incremental display of Web images as they're decompressed.
Until standards emerge for competing technologies, JPEG will continue to be the dominant compression scheme for most high-resolution color imaging applications. The JPEG specification continues to evolve through regular meetings of the X3/L3 Committee, which oversees the compression standard.
The Committee recently approved the adoption of optimized Huffman encoding, which reduced compressed file sizes by about 5%. They're now considering a proposal to replace Huffman encoding with an arithmetic-based algorithm designed to shave another 10% off JPEG compressed image files.
JPEG software is available in a variety of flavors. Royalty-free versions that originated in academia can be downloaded from a number of Web sites. Companies like Pegasus Imaging Corporation (Tampa, FL 813-875-7575) market optimized versions of JPEG targeted primarily at developers. Pegasus' products include the PIC compression libraries, with optimized versions of standard JPEG as well as Group III and Group IV.
Pegasus tools based on the PIC libraries include PICPress, an image compression, color reduction and viewing tool for Windows. PICPress works with TGA, BMP, PCX and TIFF formats and provides fully adjustable JPEG compression ratios. A variety of decompression tools such as color reduction and dithering are also available for displaying 24-bit true color images on low-bit displays with minimal color degradation.
Pegasus' IMPAC Star features a low-bit compression technology based on a process called entropy encoding. IMPAC Star uses the Pegasus IMPAC4 algorithm to provide lossless low-color image compression for images of between one and eight bits. They claim the compression ratios of this technology exceed those of competing low-bit lossless compression alternatives such as GIF and Group IV by at least 50%. Documents suited to this include fax documents, cartoons, color charts and maps.
Information about other Pegasus products including the company's PICZoom and RapidVue progressive JPEG algorithms are available at http://www.jpg.com.JPEG is also included in C-level toolkits that bundle compression with a range of related image processing functions.
LEAD Technologies (Charlotte, NC 704-332-5532), include JPEG and progressive JPEG compression in their LEADTOOLS products. LEAD says their proprietary compression format, LEAD CMP, produces smaller file sizes and better quality than standard compression formats. LEADTOOLS is available as a DLL (16- and 32-bit), and ActiveX (16- and 32-bit) and a VBX.
Accusoft (Westborough, MA 508-898-2770) includes JPEG with its ImageGear toolkit. This provides more than 200 other imaging functions in DLL and ActiveX versions.
Other toolkit vendors supporting JPEG include DFL Software (Toronto, CAN 416-487-2660), Image Alchemy Fremont, CA 510-252-0101) and Black Ice Software (Amherst, NH 603-673-1019).
In addition, Pixel Magic (Andover, MA 508-470-8830) is introducing hardware-based JPEG products. Pixel Magic says their PM-35 High Speed JPEG Codec compresses at a rate of approximately 70 Mbytes/sec -- about 100 times faster than typical JPEG software compression. The PM-35 is designed mainly for copiers and printers.
2. Fractal Compression. This technology differs from JPEG because its algorithms don't encode images at the pixel level. Instead, compression is done by analyzing and encoding repeated patterns within an image.
To understand the basics of fractal compression, put a bag of potato chips on the floor and step on it. Then open the bag and look at the fragments you've created. You'll see similarities in the fragments that allow you to create categories. A fractal compression algorithm works in much the same way. It searches the image for repeated patterns and then encodes them as equations. Once the process is complete, the pixel information about the image is discarded.
Fractal compression offers three advantages.
1. High compression ratios. Since fractal compression stores image data as equations instead of pixel information, file sizes can be much smaller. Compression ratios of more than 100:1 can be achieved without significant image degradation.
2. Fast decompression rates. When displayed, the fractal image is generated solely from the equations generated by the compression process instead of from sequentially processed pixel data. This allows fractal technology to decompress images at least twice as fast as JPEG. Fractal compression also produces smaller file sizes, which further boosts decompression rates -- and reduces storage requirements.
3. Scalability. If you want to display the compressed images at a larger scale, fractal technology interpolates the equations that represent the image. This eliminates the blockiness that typically accompanies JPEG images that have been scaled up using pixel replication.
This lets fractal images be displayed on the Web as thumbnails that load immediately, then progressively expand to full-size. In addition, fractal technology can automatically decompress images into the fixed resolution of any Web page, regardless of image size.
Fractal compression does have a downside. The complexity of the fractal encoding process means that compression rates can be slow, particularly with highly-detailed images. While you can improve performance with special hardware boards, it still lags behind JPEG in the compression speed department.
Research and development efforts are underway to improve fractal compression technology's efficiency. As fractal algorithms are refined and PC hardware platforms become more robust, the compression rates for fractal technology undoubtedly will improve.
Broad acceptance of fractals will hinge on standards. At this point, there are no formal standards efforts underway. However, fractals could quickly become a defacto standard if Microsoft and Netscape incorporate the technology into their Web browsers.
Iterated Systems (Atlanta, GA 404-264-8000) owns the patents on several of fractal compression's core components. They're also a leader in fractal technology research and development.
Iterated's recent products include Fractal Imager, a shareware tool that takes any popular 24-bit color file format (BMP, PCX, TGA, etc.) and compresses it to a fraction of its original size. A 2 MB BMP file, for example, gets replaced by a 40 KB FIF, which is Iterated's proprietary compression format. Fractal Imager versions are available for Windows 95, Windows NT and Windows 3.x (using Win32s). Check out www.iterated.com.While you're visiting their Web site, pick up a copy of their free Fractal Viewer, designed for images compressed with Fractal Imager. Fractal Viewer lets you flip, stretch and rotate an image in your Web browser. You can also copy it to the Windows clipboard and save it on your hard drive.
3. Wavelet. The most recent color compression technology is based on a mathematical concept called the wavelet. This is based on a mathematical theory (Fourier series) almost a 100 years old.
With wavelet compression, an image is divided into high-pass and low-pass image data. High-pass refers to high contrast parts of an image, as in an edge of red that meets an edge of white. Low-pass refers to areas such as a blue sky where colors modulate only slightly.
After this step is completed, wavelet compression sorts the data into a hierarchy. Image bits that aren't needed to preserve the perceptible quality of the image (selected bits within the low-pass category, in most cases) are discarded. The amount of data tossed out depends on the amount of compression you want. Compression ratios can go up to 150:1 with very little effect on the image quality.
Because the majority of high-pass bits are retained, wavelet compression often represents edges and details more efficiently than the block-based algorithms used by JPEG. Wavelet technology also uses an error-tolerant bit-stream. This means an image can be reconstructed even if data is lost while the file is being transmitted over a network.
Users of wavelet compression include the FBI, which uses the technology for its fingerprint images database. Wavelets are also used to compress the huge files created by satellite reconnaissance systems.
There are no standards for wavelet compression. Several different varieties of the technology are available, none of them compatible. However, some wavelet proponents reportedly are supporting a proposal before the JPEG committee that would replace the discrete cosine transforms currently used by JPEG compression with wavelet transforms.
Most wavelet compression is done in software with products from companies like Aware (Bedford, MA 617-276-4000). Aware was among the first companies to commercialize wavelet compression software. In 1986 Aware started working with wavelet technology when they began introducing tools for government applications. Aware has also created several software products designed for multimedia and satellite reconnaissance.
Hardware implementations of wavelet compression have also been introduced by companies such as Analog Devices (Norwood, MA 617-329-4700). These are used mainly for video editing.
Wavelet compression software is also available from Summus (Irmo, SC 803-781-5674). The Summus Still Image Compressor works with images of any size and offers several types of progressive transmission. In addition, the software can be adjusted to enhance edges and contrast, among other things. Try the Summus Web site at www.summus.com for demonstrations of the company's compression tools and information.
