Jpeg Quality

How Perceptive Are You?

Every time the topic of lossy-versus-nonlossy compression comes up on some web-mail list, it usually devolves into a pretty heated argument between several viewpoints:

In fact, the only viewpoint that doesn't seem to have any adherents left anymore are the folks who are content with ASCII art or GIFs. What's most amazing, to me, is that very few people actually do their own tests to perceptively evaluate images.

How this test works

What are we measuring here? It depends. We may be measuring how good your eyes are. We may be measuring how good your screen/display is. We may be measuring how good your graphics adapter is. We may be measuring whether or not you can tell a 90% JPEG from a 100% JPEG. In the end, my objective is simple: I just want to get you to think a little bit about image quality and perception, so that next time the lossy-versus-nonlossy compression debate comes around, you'll be armed with some opinions based on personal experience.

The test is driven by a little JavaScript program, which pre-loads all the images into memory, to make it a little harder for you to tell which file is that higher-quality one based on how long it takes to download. No information about how you scored on the test is recorded anyplace so don't worry about damage to your ego, and if you cheat you're only cheating yourself, so be my guest. If you have any doubts about the test and how it's implemented, feel free to examine the JavaScript source code (Browser -> View -> Source) and you can get the URL for the directory where the images are stored and look at the file sizes for yourself. Remember: one of the fundamental tenets of The Scientific Method is repeatability in an experiment - feel free to dissect and examine my work, and if you find any flaws in my methodology, please let me know! One known methodological flaw with this test is that I'm not using all the same quality of JPEG compression. To really do this "right" I would make you take the test 10 times, with 10 sets of images at "10 quality", 10 at "9 quality" etc. Then, I could chart at what point the average person could begin to tell the difference between images. If I were keeping the results someplace as part of a more formal study, it'd be worth doing that, but this test is for your own edification, not mine.

You will be presented with a sequence of pairs of images. One of the two images will be a lossless PNG written by photoshop, and the other will be a JPEG (quality between 7 and 12) also written by photoshop. The program will not tell you which image is which; simply click on the one that looks better to you. You will be presented with feedback as to how you perform. Clicking "Back" on your browser will not allow you to re-assess a given image, you must run the test to completion or quit out.

Click Here to Begin The Test
(It'll take a while to load since it preloads all the images, about 3MB in all. Be patient!)

Morrisdale, PA Jul 9, 2005