JPEG is an optionally 'lossy' file format; quality is rated on a 1 to 100% setting, with 60% being a reasonable standard for photos. What it does is (based on your setting) look for similar rectangular chunks in your image, average them together, then replace the similar chunks with the averaged version -- THEN compress the image, which now has more redundant data in it than the source image. At settings below 60% patchiness becomes visible in the output image, especially around text, and the image degrades because A) the patches (the averaged chunks) don't blend in smoothly with the backgrounds any more and B) the free hand you've given the computer to decide what the term 'similar' means results in a blurry-looking image. What most people don't know is that the JPEG standard permits assigning different levels of lossy to different regions of the same image, which would permit text portions of the image to be crisp while the rest was handled at a regular lossy setting. Unfortunately Photoshop doesn't support this yet in its export; you have to buy a third-party plug-in to handle that. The JPEG2000 spec is arguably the coolest yet. Among many improvements, its compression scheme stores the same image at multiple resolutions, which means better gradual image loading and more importantly, the opportunity for web and print software to pick the appropriate resolution for the task at hand. The implications for choosing render resolutions for prototyping and final renders are obvious. Unfortunately JP2K is only implemented in a few experimental science lab applications and no browser vendor is going to support a format no image editing software can export. I've skimmed the Unisys contract. While what you say may be true, I do recall the contract specifying an estimated number of licenses in a year to help define the rates -- a method they got from BMI/ASCAP and the rest of the publishing industry.