Slashdot Mirror
"Universal Jigsaw Puzzle" Hits Stores In Japan
Riktov writes "I came across this at a Tokyo toy store last week, and it's one of the coolest things I've seen in a long time. Jigazo Puzzle is a jigsaw puzzle, but you can make anything with it. It has just 300 pieces which are all just varying shades of a single color, though a few have gradations across the piece; i.e., each piece is a generic pixel. Out of the box, you can make Mona Lisa, JFK, etc, arranging it according to symbols printed on the reverse side. But here's the amazing thing: take a photo (for example, of yourself) with a cell-phone, e-mail it to the company, and they will send you back a pattern that will recreate that photo.
This article is in Japanese, but as they say, a few pictures are worth a million words. And 300 pixels are worth an infinite number of pictures."
! means factorial dude.
Impossible. Anyone who had seen the image would be a gibbering heap of insanity, and unable to tell you the pattern. At best you could hope to get enough clues to figure out the pattern yourself... but if you assembled it, you'd either off yourself or also turn into a quivering mass of human flesh.
The key here is to get someone else to assemble the image... you'd find a likely mark (some kind of paranormal investigator, for instance) and then mislead him into thinking the image he's assembling will *stop* the summoning of Cthulhu. Drop enough clues in the right places, use decoys to mislead him of your true intentions, let him be an ignorant pawn in your great game. With luck and skill, you can get him to do the dirty work for you. And the irony of him contributing to the Great Awakening by striving against it is quite delicious.
At least, that's the way I'd do it. Your way is too direct, and not worthy of true evil genius.
"Trolls they were, but filled with the evil will of their master: a fell race..." -- J.R.R. Tolkien on Olog-hai
OK, the last two are technically cheating, and all but the first option would possibly require custom code since all the example images appear to be 15x20 portraits, but a suitable algorithm probably wouldn't be that hard to figure out. I saw this on Firehose last night and worked out a few likely routines this morning, so I'd expect some custom FL/OSS code (and cheap Chinese manufactured knock-offs) to be available in fairly short order. After that the race will be on to create the largest most impressive image before the fad inevitably passes.
UNIX? They're not even circumcised! Savages!
Assuming all pieces are used, and that none of the pieces are symmetric or identical (that is, all pieces are different, and each rotation is different), then the actual number of possible images comes out to:
9*(4^300)*(300!)
where 9 is for the number of possible rectangles (1x300 up to 15x20), 4^300 accounts for the rotations of each piece, and 300! accounts for their arrangement.
The result, according to Python, works out to around 1.143*10^796, which is large, but not infinite.
I saw this at Tokyu Hands a couple days ago. Now I know what it was. The picture is only just barely similar to whatever photo you send them when you look up close. You have to view it from far away to have it appear to have the detail of the photo.
but you can make anything with it. That's like saying you can convert any picture to a 15 by 20 pixel JPEG; technically you can, but the usually the result isn't worth looking at. That said, I'm sure a lot of people will send in pr0n to convert into patterns, just to see what it looks like in ultra-low resolution monochrome.
I've abandoned my search for truth; now I'm just looking for some useful delusions.
there was a exclamation mark. 300 factorial, 300 x 299 x 298 x 297x ... x 2 x 1.
Incidentally, if anyone wants to calculate that, you'll need to use a float, and probably a double-, or quadruple-precision (YMMV) one at that.
If you're computing an integer-valued function, the result should be an integer. In Python, which uses arbitrary precision integers by default, it's as simple as:
>>> import math
>>> math.factorial(300)
306057512... [truncated to get past the lameness filter]
JPEG chunks an image into 8x8 blocks. An overcompressed JPEG contains so little information per block that the blocks devolve into simple gradient patterns (try this yourself with a grayscale image: save it with a quality near "0" and you will see the individual blocks clearly). If you think about it a bit, this makes sense: the block is being approximated by a combination of a small number of cosine waves (in the limit, it's a single wave along each image dimension), so the result is a gradient, because most of the coefficients have been thrown out by compression.
In this sense, the puzzle pieces can be thought of as representing these simple block patterns. With a 15x20 rectangle of pieces, by JPEG standards, this is essentially an overcompressed 120x160 image. You'll note that if you take your overcompressed JPEG and scale it down to around 25% (30x40), then, provided the original image shows only a single subject, it should still be reasonably recognizable, because the human visual system patches together the pieces to produce a coherent image, even if it is highly distorted.
where 9 is for the number of possible rectangles
This also assumes that you're building a rectangle.
and that all your pieces are attached, if you allow arbitrary separation then the number of combinations is infinite.
The question of whether a computer can think is no more interesting than the question of whether a submarine can swim.
Mod parent up.
It's the gradients on the pieces, and the principles of human vision that JPEG takes advantage of, that give this puzzle its cool effect, creating the appearance of a much higher resolution than the 15x20 "pixels" everyone else is referring to.
You can't make a (easily) recognizable Mona Lisa in 15x20 pixels. You can in 15x20 cosine gradients.
Your mind is clear / The things that you fear / Will fade with how much you / Believe what you hear
306 057 512 216 440 636 035 370 461 297 268 629 388 588 804 173 576 999 416 776 741 259 476 533 176 716 867 465 515 291 422 477 573 349 939 147 888 701 726 368 864 263 907 759 003 154 226 842 927 906 974 559 841 225 476 930 271 954 604 008 012 215 776 252 176 854 255 965 356 903 506 788 725 264 321 896 264 299 365 204 576 448 830 388 909 753 943 489 625 436 053 225 980 776 521 270 822 437 639 449 120 128 678 675 368 305 712 293 681 943 649 956 460 498 166 450 227 716 500 185 176 546 469 340 112 226 034 729 724 066 333 258 583 506 870 150 169 794 168 850 353 752 137 554 910 289 126 407 157 154 830 282 284 937 952 636 580 145 235 233 156 936 482 233 436 799 254 594 095 276 820 608 062 232 812 387 383 880 817 049 600 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000
Stylish sheet to fix many problems in Slashdot's D3: https://gist.github.com/801524
Wikipedia explains it fairly well, though it is still a bit technical. The Huffman coding details are not important to the main idea; what is important is the concept of using the Discrete Cosine Transform (DCT) and a subsequent quantization step to reduce the precision, and therefore size, of the high-frequency components.
At extreme compression levels, only a few of the patterns in this image will end up with non-zero coefficients: specifically, those which are low-frequency (in the top left corner). You'll notice that these also happen to resemble gradients, which is what I meant when I said that the 8x8 blocks devolve into simple gradients.
FWIW it says take a photo with your cell phone, send it in and the response arrives.
They have variations in sepia, wine red, midnight blue, and a puzzle game -- but all are sold out.
By the way JIGAZO actually is how you read the three character name, which means "your own image" or maybe "self portrait image".
Here is what the separate white panel looks like:
here
And here is a blog, showing the process: "I made the Jigazo puzzle!" ;)
I'm not going to translate it entirely, but he says the pieces are nice and thick, and well formed. Also there is a guide image on the back of each piece and also, a light version of it is shown on the front of the piece as well so it's easy to understand.
It comes out looking quite good, the key is to look at it as if looking far away, with squinty eyes.
It took him 90 minutes to make his first one, then the second time he got faster.
Translation..
Use the 300 pieces in the box and you can make anybody's face.
First in the World, a jigsaw puzzle that can do your face!
Jigazo Puzzle
Can you believe it?!
All the sample images shown here were made with the same 300 pieces.
In order to make your own face...
When you send a photo by cell phone email a response image will be sent back immediately.
Just put the pieces together the way it shows and voila!
By repositioning the same 300 pieces you can make images as varied as these!
1. Send your image to the email address printed in the included instruction manual
2. A response image with the answer arrives in your cell phone
3. When you line up the 300 pieces according to that, then your face is complete. You can use anyone's face!
(small print)
Sending your photo, and receiving the response image will incur packet communications charges according to your cell phone service provider's contract.
If you are going to be using it repeatedly, we recommend you use your service provider's fixed price unlimited packet plan.
Note that the creation of the response image by our company is without cost no matter how many times you use it.
Set contents
300 piece puzzle. Color of pieces differs depending on the set type.
Piece set-up tray
Starch adhesive and applicator spatula (or something like that)
Manual. Includes instruction images for the Mona Lisa, Girl with Pearl Earrings, Natsume Soseki (the author), Kennedy (JFK), President Obama, Beethoven's face. You can immediately start playing with these faces without using a cell phone.
Other things you need:
A cell phone with camera functionality, from NTT DoCoMo, AU, Softbank Mobile (Disney Mobile and iPhone also included).
* When displaying your finished puzzle, please use separately sold 300 piece (white) panel specially for JIGAZO puzzles.
Price is 1890 yen, a bit over 20 bucks.
JIGAZO FAQ
Q1. When I send in a portrait photo with my cell phone, do you mail me back a finished puzzle?
A. No, that is incorrect. Your face is created only out of the 300 pieces in the box.
Q2. How do you do that?
A. There are 300 pieces in the box with slightly different tones.A program finds the tones closest to those of your face and sends back to your cell phone a response image specially for your face.
Q3. How quickly does the response image get sent back?
A. It depends on the state of the communications network, but in about 10 seconds it should be sent back and then you can immediately start playing.
Q4. How do you put the puzzle together?
A. Each of the pieces has a hint image on it, so you position the pieces as shown by the response animation. Enjoy watching your face gradually start to appear.
Q5. Can I only put the pieces together once?
A. No, you can redo it any number of times, and make anyone's face.You can pl