3 ms·
My results with libvips have been way more encouraging than this post suggests with MozJPEG and the WebP reference encoder! Here's a MEGA folder with my exampl
by Lammy 6y ago
My results with libvips have been way more encouraging than this post suggests with MozJPEG and the WebP reference encoder!
Here's a MEGA folder with my example files so you can compare for yourself. Sorry I know this host might be blocked in many places, but every image host I tried (imgur, postimg) recompressed my uploaded files negating the fairness of the test. If anyone has a suggestion for a better host I'm happy to re-upload.
https://mega.nz/folder/JsUXWIqI#xYIohBdiVKrGYTi-BeENEA https://mega.nz/folder/JsUXWIqI#xYIohBdiVKrGYTi-BeENEA
e: another host: http://www.filedropper.com/libvipswebpvsjpeg http://www.filedropper.com/libvipswebpvsjpeg
My random sample is a black-and-white JPEG photograph vs its WebP equivalent. The results are in line with my results for many other images of different types, like other black-and-white photos, color photos, clean computer graphics, screenshots of UIs and media/games, etc etc. The original image for this test can be found here: https://en.wikipedia.org/wiki/Sand_Hill_Road https://en.wikipedia.org/wiki/Sand_Hill_Road
Both test images were output by libvips after loading the original JPEG. The libvips-generated JPEG is 1.22mb, smaller than the 1.26MB original, but the libvips-generated WebP is 944KB! I included two sceenshots of each libvips-generated version opened in a same-size window too, both much smaller than the full res. Tabbing between the two on my PC does not visibly change on screen at all.
I don't entirely enjoy WebP just because the software support is still spotty, but I haven't been able to argue with the compression benefits!