JPEG of the New Millennium

Not too many people know that JPEG stands for Joint Photographic Experts Group. But what they do know is that it is a popular image-compression technique. It reduces the size of an image file by applying certain compression algorithms, resulting in an image file with a .jpg extension. In such compression, however, bits and pieces of image information (usually which cannot be perceived by human eyes) are lost, resulting in ‘lossy’ compression. In this technique, therefore, the higher the compression, the lower the image quality. Though there’s also a lossless JPEG compression, called JPEG LS, it is not very popular. 

The new avatar of JPEG, namely JPEG 2000, promises improvements over JPEG in compression, image quality and performance. It also adds some new features for the current and next-generation applications. Let’s look at what the main
offerings of JPEG 2000 are and how they compare.

Better compression technique

JPEG compression is lossy because it uses DCT (Discrete Cosine Transformation), which separates the low frequency and high frequency information in the images. Then depending on the desired size (or quality) of the image, more and more of higher frequency information are discarded.

JPEG 2000, on the other hand, uses DWT (Discrete Wavelet Transform) compression. DWT is more effective over DCT as it preserves the frequency information in the image leading to lossless compression. This is similar to the popular zip (WinZip may sound more familiar) compression, which compresses a file (by often reducing its size) and upon decompression yields the original file without any loss of information. 

JPEG 2000 supports both lossless and lossy compression. While lossless compression is suited for applications like medical analysis, scientific research, lossy compression is required to produce images for the Web, wireless and embedded devices.

Low artifacts

In DCT, the input image is divided into compressed blocks, which result into blocking effects or blocking artifacts. DWT, on the other hand, works on the entire image. In DWT, the compressed image is stored as wavelets (mathematical functions) as opposed to pixel blocks in DCT. This makes DWT more storage efficient.

Single image at different resolutions

With JPEG 2000, an image can be encoded in layers at different resolutions. While the lower-resolution layers can be used for thumbnail views, the higher-resolution ones can be used for detailed full-screen views. You can determine the resolution or quality at which you want to see an image. This is somewhat similar to Flash animations where you can select the animation quality by right clicking on it and selecting the quality level. When such images are used on the Web, you can first download the lower-resolution layers and then the higher-resolution ones. Once you are satisfied with the image resolution, you can stop further downloads.

With JPEG 2000, you can also encode some parts of an image at higher resolution. For example, in an image of a circuitry board, you can show a particular onboard chip at a higher resolution so that the specifications written on it are clearly visible. This type of encoding is called ROI (Region of Interest) coding. 

XML metadata

Besides image data, JPEG 2000 allows the inclusion of metadata in an image file. This metadata may contain more information on the image data, like colors, tones, author information, IPR (Intellectual Property Rights) information and the details of the device used to capture the image. What’s more, the metadata can be extended to accommodate additional information. The metadata is in XML so it can be easily extracted within various applications using/coding XML parsers. 

Shekhar Govindarajan