Hacker News new | ask | show | jobs
by vytis 5400 days ago
Well the quality seems quite a bit worse for the picture I tried. Original: http://postimage.org/image/94ol50ck/, compressed: http://postimage.org/image/14712qgjo/ The resulting image has way less detail in higher frequencies.
2 comments

looking at the file names I think this is the compressed version http://s3.postimage.org/2qgfsekg7/Front_mini.jpg

I'd like to run it through errorlevelanalysis.com but their service seem to be down.

I suspect it tries to raise the brightness and sharpness of certain pixels before compressing it further so "important" elements stay in detail. What was done to your image though looks nothing like the quality of their examples for the ship, etc.

Looks like the compressed version is a bit sharper to me.
Sorry, I pasted the links in the wrong order. The original is this one: http://postimage.org/image/14712qgjo/ and the compressed image: http://postimage.org/image/94ol50ck/ File size was reduced from 4.4mb to 1.2mb
If the compressed version is apparently sharper than the original file, that's an artifact surely? It might be pleasing in some contexts but it's not an accurate representation of the original file in that aspect.
I mean, vytis claimed that

> The resulting image has way less detail in higher frequencies.

But the compressed version actually exposes more detail.

But the compressed version actually exposes more detail.

No. It really doesn't.

It adds a couple of visual filters (brightening, sharpening) before recompressing, but these don't "expose more detail", but rather all three steps actually introduce additional errors. Errors that trick they eye into seeing a "better" image, but errors nonetheless.

JPEGmini does not apply any filters on the photo, no pre processing whatsoever. JPEGmini went through BT.500 certification, the result was that given 2 images 1. source 2. recompressed, the tester could not tell which is the source and which is the recompressed. Enjoy.
I'm not sure what BT.500 certification is, but I can certainly tell the difference on some (not all) images, when recomposed.

Are you attached to JPEGmini? If so, would you be willing to bet that I cannot tell if a photo of my choice has gone through your system and been compressed?

I'm on a 3G connection today, so relied on this thread for before/after images. Unfortunately, it seems that the original poster had labelled his photos the wrong way around...
Think he might have got the links the wrong way round