Lossy vs. lossless

Sunday 16 August, 2009 16:39

Note: This post was partly inspired by a discussion at the AV Science Forum. To view the original conversation and “test”, take a gander here.

One issue that rears its head again and again with regard to the posting of screen captures from Blu-ray Discs is the use of a compressed file format such as JPEG as opposed to an uncompressed format such as .BMP or a compressed but lossless format like PNG. I’ve been posting full resolution screen captures for some time now, and, like a number of other sites doing the same thing, have always used JPEGs. As a result, the reliability of such captures have been called into question on occasions.

This is not unreasonable. I lean towards scepticism in many aspects of life, so the notion that people would be wary of trusting compressed images to tell the whole story is one I can fully understand. However, I thought I’d take a little time to explain why I use compressed images, and more importantly why I believe the images I post are an accurate reflection of the source material, and are not adversely affected by the aforementioned compression.

Fundamentally, the issue is one of space and bandwidth. A PNG image with a resolution of 1920x1080 can have a size of upwards of 3.5 MB, depending on the complexity of the image. (An uncompressed bitmap image with the same resolution will always occupy 5.93 MB in 24-bit colour.) A JPEG, saved with minimal compression (a value of “12” in Adobe Photoshop), takes up considerably less space, ranging in my experience from around 500 KB to around 1.4 MB. My web host provides me with 10 GB of disk space and a monthly bandwidth limit of 100 GB, which sounds like a lot but in reality is quickly eaten up by having vast numbers of large images stored and repeatedly accessed, particularly when you take into account leeching. Storing all my images as PNGs simply isn’t feasible. I can either use compression or post substantially fewer images. (The possibility of hosting my images on free image sharing services like Photobucket has been suggested to me. However, in my experience these are always considerably slower than my host, which would only frustrate those viewing the images.)

Unlike some sites that I could but won’t mention, it has always been my policy to use minimal compression. As stated above, this means saving my JPEGs in Photoshop with a “quality” setting of “12”. Time and again, my experience has been that this results in no noticeable degradation of even the most complex images. Zoom in to 3,200% or something equally ludicrous, and you may start to see the odd individual pixel being slightly displaced. In real world terms, this is nothing. Such differences aren’t visible at 100%, or even 200% or indeed 600% size, and therefore in my opinion have absolutely no bearing on the purpose of these captures, which is to provide as close as possible to an accurate impression of what a transfer looks like without actually viewing it in motion.* It’s entirely true to say that they are not a carbon copy of the original frame down to the individual pixel. However, I contend that such differences are, from our point of view, essentially meaningless. We’re not looking at these still frames under microscopes… and if we are, we should probably get out more.

At the suggestion of FoxyMulder (who posts on both AVS and Land of Whimsy), I devised a little test, posting two copies of the same capture of the same frame from the same film. Both are PNGs. However, one was first saved as a compressed JPG (quality of “12” in Photoshop), then saved again as a PNG. I’ve provided three examples below. In each case, one was saved lossily and the other wasn’t. Can anyone tell me, without cheating (if you don’t know how I’m not going to tell you), which is which?

#1. Monsters, Inc.

Monsters, Inc.
Image 1 / Image 2

#2. Franklyn

Image 1 / Image 2

#3. The Dark Knight

The Dark Knight
Image 1 / Image 2

Post your answers by Tuesday and I’ll post the correct answers, along with how many people were right.

* I’ve said it before and I’ll say it again: there’s no substitute for actually viewing a disc for yourself. You can tell a lot from screen captures, but unfortunately they don’t and never will be conclusive. For one thing, while they display the grain structure of an individual frame, they don’t show its cumulative effect in motion, which can create quite a different look.