| |  |
BitJazz Developer Support
PhotoJazz Development FAQ
Competition
What about JPEG2000?
What about JPEG?
 What about JPEG?
JPEG is an inherently lossy compression algorithm, even at maximum quality.
For example, the peak error for Kodak's standard test images using Photoshop's JPEG at maximum (10) quality ranges from ±6 to ±9, or around 3 bits.
For an analysis of JPEG artefacts, see the page on JPEG lossiness.
By destroying perceptually relatively unimportant information, JPEG is capable of compressing an image much further than any lossless compressor. This makes JPEG ideal for broadcast-quality images that will not be processed further after conversion to JPEG format.
BitJazz is a lossless compressor, and therefore cannot use any of the lossy processes employed by JPEG: color-space conversion, chroma decimation, discrete cosine transforms, and spectral truncation.
Moreover, JPEG does not support duotone, CIE-Lab, or multichannel image modes, nor does it support any alpha or spot-color channels, and it has no facilities for data integrity verification.
top | next
What about JPEG2000? BitJazz is as powerful an image compressor today as JPEG2000 is expected to be two years from today. By the time JPEG2000 is finalized, BitJazz will be much more powerful. However, BitJazz and JPEG2000 will eventually converge along several paths.
JPEG2000 is scheduled to become an international standard in November 2000, and international standards organizations are well-known for the thoroughness and hence slowness of their deliberations, as they properly should be.
The submission deadline for compression algorithms for JPEG2000 was September 1997, and since BitJazz Inc. did not even begin its research program in lossless image compression until that very month, we were of course unable to contribute our algorithm for consideration. Despite our late start, however, we have moved quickly and already have a superior product on the market.
JPEG2000 will most likely use wavelet transforms, which have been amply demonstrated to perform better than cosine transforms (as in the current JPEG) for lossy compression. For lossless compression, however, wavelets offer little or no improvement over other methods. The JPEG committee has actually already advanced two prior lossless compression algorithms. The first, included as part of the original JPEG specification, used DPCM (differential pulse-code modulation), and performed so poorly that no popular software ever implemented it. The second, JPEG-LS, has not yet been released.
DPCM, Ziv-Lempel, wavelets, fractals, and cosine transforms are all quite mature, having been around for 10 to 20 years, and little additional improvement can be expected from them. Indeed, it is too late for any improvements to be introduced into candidate JPEG2000 algorithms, which are currently scheduled to be in the ballot stage. BitJazz's condensation technology, in contrast, is in its infancy, and still has great potential for improvement.
However, one of the many items on the JPEG2000 wish-list is an open architecture that permits the inclusion of other compression algorithms. In fact, another desired feature is that the current JPEG algorithm be supported in this fashion for backward compatibility. Once the JPEG2000 committee begins accepting new compression algorithms, we will of course be very interested in submitting our algorithm.
On the other hand, the PhotoJazz file format also supports multiple compression algorithms, so we may well end up adding whatever algorithm JPEG2000 settles on to BitJazz. And most of the features on JPEG2000's wish-list that PhotoJazz does not yet support are features we are planning to add also.
In sum, PhotoJazz and JPEG2000 are likely to converge eventually in several different ways. In the meantime, PhotoJazz is available now.
previous | top
|
|
|
