JPEG 2000 fixed JPEG's block artifacts with wavelets, but lost the format war anyway

By the late 1990s, JPEG's block-based discrete cosine transform was showing its limits: at high compression, images developed visible 8x8 block artifacts, and the format could not do lossless or progressive decoding well. The JPEG committee started a successor in 1997, and in 2000 standardized JPEG 2000, built on whole-image wavelet transforms.

The wavelet approach changes the mechanism, not just the numbers. Artifacts spread gracefully across the image instead of locking into blocks; one file can be decoded progressively, so a viewer sees a rough version almost instantly; and both lossless and lossy modes live in the same standard. JPEG 2000 also has explicit support for regions of interest and no royalty burden for the baseline decoder.

It was, by most technical measures, better. Yet JPEG 2000 never displaced JPEG: the old format was already everywhere, encoders and hardware were built around it, and the marginal benefit was not enough to move an entrenched ecosystem.

Wavelets remove blocks; one file, lossy and lossless

Wavelet compression spreads error gracefully, eliminating the visible block boundaries of DCT-based JPEG.

Progressive decoding in one file is genuinely useful for slow connections and large imagery.

One standard covering lossless and lossy removes the need to choose a separate format per use case.

The failure was not technical: legacy encoders, cameras and browsers made JPEG the compatibility floor.

A technically superior mechanism can still lose if the incumbent's ecosystem—encoders, browsers, cameras, patents—is what actually defines compatibility.

JPEG 2000 found niches that needed its strengths—digital cinema, medical imaging, archival scanning, and some GIS and broadcast workflows—but consumer imaging stayed on JPEG. The format war lesson became a case study: a better mechanism does not beat an installed base by itself.