Yeah, I'm very sorry about that. Mixed the files up when uploading and left the testing one up by mistake :)
May I know how you obtained such a result by the way?
I used Python. You just need to install Pillow library and put ico files with this in the same folder and run it. :)
At first I thought only different pixels are important, so I painted them in two colors. One for lower and one for higher values of blue. BWT, why blue only? After seeing this I changed code a bit (commented part) to produce final image.
But that's just the blue component comparison... Oh damn, I guess I made a small mistake again :D It shouldn't be blue only by design, you're right, it's too noticeable. I'll look into it now.
Thank you :) Well... If you're familiar with some utilities it's an easy one-liner with ImageMagick (which is really powerful by the way) not requiring any additional coding:
compare -highlight-color White -lowlight-color Black ecc200.ico favicon.ico out.png
Since other people are sharing their secrets, I'll mention that I created an OpenCV program using C++ that compares pixels and sees if they're EXACTLY the same.
I was tempted to use Python since it'd be easier, but I figured it was a nice chance to get to know gimp a bit. It turned out to be a lot of work. Specially since I needed to deal with the alpha layer too which gets special treatment. Ended up manually splitting each channel, comparing them, putting them back together.
Even just how to get the differences to show up was difficult due to using the LSB. From googling it seems like Photoshop has a lot of more layer types that would've been useful, like threshold. But gimp's threshold tool only worked on the current layer, so it took me a while to figure out to merge layers, and I wanted to avoid doing that so I didn't lose data as I was still figuring things out at the time, etc. Of course, gimp has access to scripting and such like some other posters mentioned. It does seem like photoshop has many more ways to do things in a simple way, like I'd read before. Or maybe there is a better way to do it with gimp without getting into scripting.
Of course, a simple (python) script would've sufficed. Loop through 4 byte blocks in each file. If the difference isn't 0, set the pixel to FFFFFF(FF/00? don't remember which alpha value is opaque) but that's no fun. :P
First! :D
Comment has been collapsed.
Congratulations :)
Comment has been collapsed.
Thanks. :)
First attempt was... Well you know: click
Comment has been collapsed.
Yeah, I'm very sorry about that. Mixed the files up when uploading and left the testing one up by mistake :)
May I know how you obtained such a result by the way?
Comment has been collapsed.
I used Python. You just need to install Pillow library and put ico files with this in the same folder and run it. :)
At first I thought only different pixels are important, so I painted them in two colors. One for lower and one for higher values of blue. BWT, why blue only? After seeing this I changed code a bit (commented part) to produce final image.
Comment has been collapsed.
But that's just the blue component comparison... Oh damn, I guess I made a small mistake again :D It shouldn't be blue only by design, you're right, it's too noticeable. I'll look into it now.
Comment has been collapsed.
I saw pattern (blue only) manually in GIMP, so I didn' check it. :D
Comment has been collapsed.
Damn it X) It should be random to be (mostly) invisible in imaging software of course. I'm fixing it.
Comment has been collapsed.
Fixed, check it out again if you want :)
Comment has been collapsed.
Edited code (commented part). Works well. :D
Comment has been collapsed.
I think it was the last bug in there :)
Comment has been collapsed.
I even know how to solve it in GIMP, but it works only for R, G and B. Alpha channel is treated differently.
It was great puzzle. Can't wait for next one. :D
Comment has been collapsed.
Thank you :) Well... If you're familiar with some utilities it's an easy one-liner with ImageMagick (which is really powerful by the way) not requiring any additional coding:
compare -highlight-color White -lowlight-color Black ecc200.ico favicon.ico out.pngKnow your 'quality optics' ;)
Comment has been collapsed.
Thanks.
Comment has been collapsed.
Hope you liked it ;)
Comment has been collapsed.
Since other people are sharing their secrets, I'll mention that I created an OpenCV program using C++ that compares pixels and sees if they're EXACTLY the same.
Comment has been collapsed.
Hi ^_^
My solution is basically the same as anadius' only in Java.
Thanks alamarjan :)
Comment has been collapsed.
Sure thing :)
Comment has been collapsed.
Nice puzzle, thanks :)
At first, I wanted to compare images in Python, but then I thought, there already is a tool for that - ImageMagick.
Then, I used online scanner to decode DataMatrix :)
Comment has been collapsed.
Hello! Nice to see you here :)
Comment has been collapsed.
Python's PIL failed me :(((. Used GIMP's script for 'comparing images' - worked wonders. Now I'm gonna find out what's wrong with PIL...
Comment has been collapsed.
Congratulations chour!
Comment has been collapsed.
I was tempted to use Python since it'd be easier, but I figured it was a nice chance to get to know gimp a bit. It turned out to be a lot of work. Specially since I needed to deal with the alpha layer too which gets special treatment. Ended up manually splitting each channel, comparing them, putting them back together.
Even just how to get the differences to show up was difficult due to using the LSB. From googling it seems like Photoshop has a lot of more layer types that would've been useful, like threshold. But gimp's threshold tool only worked on the current layer, so it took me a while to figure out to merge layers, and I wanted to avoid doing that so I didn't lose data as I was still figuring things out at the time, etc. Of course, gimp has access to scripting and such like some other posters mentioned. It does seem like photoshop has many more ways to do things in a simple way, like I'd read before. Or maybe there is a better way to do it with gimp without getting into scripting.
Of course, a simple (python) script would've sufficed. Loop through 4 byte blocks in each file. If the difference isn't 0, set the pixel to FFFFFF(FF/00? don't remember which alpha value is opaque) but that's no fun. :P
Comment has been collapsed.
Look at the imagemagick command mentioned earlier :)
Comment has been collapsed.
Thank you, this was interesting puzzle)
Comment has been collapsed.
I've just discovered SteamGifts... and its forum... which contains puzzles... this could be either very fun or very distracting.
I solved it once I thought to actually pay attention to the file name. Ah, ECC 200 is a barcode... well that explains all the black pixels. Haha.
Here's my really short Mathematica solution:
a = First@Import["http://obscuregamers.com/files/ecc200.ico", "Data"];
b = First@Import["http://www.steamgifts.com/favicon.ico", "Data"];
BarcodeRecognize@Image@Boole@Map[# != {0, 0, 0, 0} &, BitXor[a, b], {2}]
Comment has been collapsed.
Or an even simpler solution:
a = First@Import["http://obscuregamers.com/files/ecc200.ico"];
b = First@Import["http://www.steamgifts.com/favicon.ico"];
BarcodeRecognize@ImageApply[Max, ImageDifference[a, b]]
Comment has been collapsed.
That's good :) Welcome!
Comment has been collapsed.