JPEGmini vs Gimp JPEG vs WebP


Recently there is a lot of buzz about JPEGmini. They pretend to reduce JPEG file size up to 5 times without loss of quality and artifacts:

“original JPEG photo and the recompressed JPEGmini photo cannot be distinguished by average users when displayed at 100% zoom on a computer monitor”

Well, it's not true. The loss of quality in JPEGmini is comparable to traditional methods.

I tried to re-compress this image, the original JPEG was 1024x680, 350.1 KB:

Original JPEG 1024x680

JPEGmini produced a 102.9 KB file, 3.5 times smaller, but it lost a lot of sharpness, lost some details in low-contrast areas and acquired some artifacts along the edges. The image quality was clearly inferior to the original when compared directly.

I tried to achieve the same file size by re-compressing the image in Gimp. Quality=85% and smoothing=0.22 gave me a 102.7 KB file. Guess what, it looked very similar to JPEGmini. It had slightly more artifacts, but preserved more detail in low-contrast areas.

Finally, I tried to re-compress the image to WebP, a new image format which is much more effective than JPEG. At quality=88 I got a file size of 92.9 KB, less than that of JPEGmini. WebP was visibly sharper than JPEGmini, but there was visible blocking in low-cotrast areas.

As an example, look at all four samples one after another in this animated GIF, crop at 100%:

JPEGmini vs Gimp vs WebP @ 100%

And the same crop, zoomed at 300%:

JPEGmini vs Gimp vs WebP @ 300%

JPEGmini's choice of quality is questionable.

Corollary 1: JPEGmini is not magic.

Corollary 2: Neither is WebP.