^ Will depend a bit on what deinterlacer you use as well.
Tally:
A: 10
B: 2
C: 1
Tally from PSP-Hacks:
A: 8 (including Kuu/Zero's vote)
B: 1
C: 5
Unsure: 5
As most of you have guessed, the correct answer is A.
Here's the original 1080p image:
http://zingaburga.com/upload/aorig.png - overlay the two to reveal certain details not present in A.
Results were probably better than I had expected. But perhaps not so good considering the numbers. 1080p has 460,800 more pixels than 720p, and over twice the area (2.25x). A significant increase in detail is somewhat expected, however this experiment shows some doubt casted between the two in terms of perceivability.
I'm not really sure how much you criticised the images before you made the decision, but if I rely on what I heard in IRC, you probably can't tell the difference at a first simple glance. It even may withstand a quick degree of scrutiny, depending on how sensitive you are, but should be obvious after heavy scrutiny.
Considering that most of you deal a fair bit in graphics/video/gaming, you should be more sensitive to these small defects (eg if you're trying to determine whether an image is Shopped :P or playing around with game quality settings to see the difference) than the average viewer. Another thing is that here, you can compare detail levels, whereas if you just got a 720p clip that was upscaled (without your knowledge) to 1080p, for example, you don't really have much base for comparison.
Problems with this experiment
This really was meant to be a quick test - obviously there's a number of issues, such as:
- Still image VS video - being able to tell the difference between still images doesn't necessarily mean the same for video. With a video, for example, you can't really scrutinise images as much unless you pause the video.
- Compression artefacts - upscaling a 720p video will upscale any compression artefacts, whereas this downscale » upscale probably won't show them up so obviously. Another point, scenes with more detail are likely to have a higher quantiser, and thus, less detail regardless of resolution.
- Scene complexity and differing scenes used
- Impractical resampler - to obtain image A, I grabbed the original, downscale to 720p using S-Spline XL (Downscale preset) and upscale to 1080p using S-Spline XL (Graphics preset). Your media player is unlikely to have such a complex resampler. The most widely used appears to be Bilinear, which will make the image look like the following: http://zingaburga.com/upload/abilin.png
- You have reference images rather than being just presented with an image to determine whether it's really 720p or 1080p (probably what you'd usually encounter)
- etc etc
One of the original ideas I was having was whether it was possible for someone to upload a "1080p" movie which is really a 720p upscale using a quality resampler,
and no-one actually notice it.
Thanks for your participation :)