In the end, JPEG2000 is not even good at compressing images.
It uses wavelets which are overfitting the PSNR metric. In purely numeric benchmarks (without humans judging the images) it makes JPEG2000 seem way better than it really is. Wavelets may look impressive when you compress an image to blurry death, but that's a niche use-case. It falls apart in the more relevant high and nearly-lossless quality ranges. It handles sharp edges very poorly, and struggles to preserve textures of things like skin or wood — either everything looks like plastic, or you get poor compression.
DCT-based codecs can generate realistically-looking textures with fewer bits. For web uses even de-blocking filters of WebP and AVIF are not that useful, because people want to use qualities at which blocking isn't a problem in the first place.
It uses wavelets which are overfitting the PSNR metric. In purely numeric benchmarks (without humans judging the images) it makes JPEG2000 seem way better than it really is. Wavelets may look impressive when you compress an image to blurry death, but that's a niche use-case. It falls apart in the more relevant high and nearly-lossless quality ranges. It handles sharp edges very poorly, and struggles to preserve textures of things like skin or wood — either everything looks like plastic, or you get poor compression.
DCT-based codecs can generate realistically-looking textures with fewer bits. For web uses even de-blocking filters of WebP and AVIF are not that useful, because people want to use qualities at which blocking isn't a problem in the first place.