LGP Announces New Competition
Time Doctor writes "Linux Game Publishing announced its new game competition today, wherein an image relating to the game is revealed one pixel a second and competitors can attempt to be the first to guess it. Winner gets the first copy of the game, and the unofficial award of having way too much time on their hands to sit around waiting for pixels to change."
they are putting a dynamically generated 1085x814 image that changes once a second on to a site where we here at slashdot are going to check it out repeatedly? That doesn't appear to be a very bright idea.
Couldn't they provide something else than a compressed jpeg full of jpeg artifact (zoom the large picture, you will see a the image is composed of 8x8 block of seemingly random pixels)
How is one supposed to know what the hell is in there if the jpeg compression moves the changed pixels around?
JPEG image can't be revealed one pixel at a time. JPEG image consists of 16x16 MCU (Minimal Coding Units) encoded with DCT and high harmonics discarded (actually, there's more to this). Changing one pixel before encoding changes the whole 16x16 square.
You won't be able to see patterns very well at all unless they post the original picture so you can do a diff. Does anyone have a copy of the original picture? [Or do you know where a link to it is?]
I think this is a pretty stupid way of doing it. They should have just done it from a blank image. This just gives people who know the original image an advantage.
I'm sorry. The number you have reached is imaginary. Please rotate your phone 90 degrees and try again.
The full image is 1280x960. 14 days for the full image.
That what was all this school was for... to teach us how to solve our own problems. -- janeowit
A real geek whould get the first image and then xor it with the one sometime later.
So picking out all the random dots and leaving all the information dots, makes guessing a lot easier.
Greets
There are no stupid questions, Just a lot of inquisitive idiots. (from a good friend)
The image looks like it might be largely symmetrical. It'd be worth combining the two halves, though I'm not sure what procedure would give the best results.
Maybe average the colours in 16x16 blocks (does that eliminate the jpeg noise?) then average the two halves. Or just check for pixels that are the same shade on each side, this throws out most of the data but even more of the noise.
I quit!
Anyone got the original, very first image they posted?
There's got to be a way to at least make the challenge easier. All the random pixels just confuse my visual cortex, so blacking them out, leaving only the pixels already revealed (about 45000 by the time I post this) would certainly make the job easier.
Assorted stuff I do sometimes: Lemuria.org