[magicalhippo]

Dithering isn't a compression algorithm[1]. It's an algorithm that adds noise to effectively increase bitdepth when downsampling.

In the case of images, it makes color-reduced versions not look terrible.

Adding noise is almost always a bad thing for compression. The undithered images would compress much better.

[1]: https://en.wikipedia.org/wiki/Dither

[WithinReason]

It is when combined with quantisation, and the article specifically refers to it as such.

[rjmunro]

If added noise is the aesthetic effect you are going for, it's probably better to ship a low quality (<50) JPEG of the original image and add the "dithering" in the client by overlaying noise there.

[hunter2_]

You can't add the noise after quantization, it needs to be done at the same time as quantization, because otherwise you are losing the information needed to intelligently feather the edges.

Think of it this way. You start out with a series of numbers between 0 and 100. Your job is to represent this series as best as possible within a range of just 0 to 10. Without dither, you would just round each original number to its closest multiple of 10; all your 31's become 3 and all your 34's become 3. With dither, nearly all of your 31's become 3 and many of your 34's become 3 but nearly half of them become 4.

Without dither: 31, 31, 34, 34 becomes 3 3 3 3.

With dither: 31, 31, 34, 34 might become 3 3 3 4 on a typical run.

You absolutely cannot calculate 3 3 3 4 based on 3 3 3 3. You need the original full set of information in order to calculate 3 3 3 4.

Now add on the fact that it's not just random noise that makes this work. Neighboring pixels influence whether to round up or round down. You want that influence to come from original high-depth data, not already-rounded data.

[marginalia_nu]

Dithering is fairly expensive operation. Please don't do it client-side each time a resource is fetched.

[WithinReason]

It's much cheaper than decompressing a JPEG

[marginalia_nu]

Compressing, perhaps. But decompressing? Not really.

[WithinReason]

Floyd-steinberg dithering is at least an order of magnitude cheaper than JPEG decompression

[marginalia_nu]

Maybe if all else is equal, but let's not forget you'd be doing the dithering in javascript, and the JPEG decompression is implemented in a low level language, in practice probably hand written assembly (as performance-critical compression code often is).