The RPMs should have the rubbersheet stuff inside already (complied into the library).
I've added an extra download for the binary for just the plugin, in case you build from source. Check out the vips-7.8 download page again.
There are some extras in the "extras" bit at the top of the website. nip has only been launched properly for 6 months, so there's no public repository yet.
Hi, I'm the VIPS maintainer. Thanks for the plug Yarn:)
It has a rubbersheet tool. You make a target image (a black and white chessboard works well), print the target out, take a picture of it (you'll need to pick a particular zoom setting to calibrate), and feed the original (undistored) image and the (distorted) photo into the transformation estimator.
The estimator uses gradient analysis to iteratively discover a 0/1/2nd order transform from the distorted image back to the original. It won't work so well at the image edges (you'll need to crop the corrected image down a bit), but mostly it works pretty well. I use it on 10,000 by 10,000 pixel images without problems. For pincushion/barell distortions it should get you a fixed image to less than a pixel error.
Downside: VIPS is GPL and the transform stuff was done by a friend of mine as part of his PhD. He doesn't want to GPL it, so the transform stuff is distributed as a binary plugin. The windows and suse binaries on the website include it.
You're right, but of course the service is currently restricted to US-resident, Mac-owning, OSX-installed broadband users. When they open the user base out a bit there should be a very significant jump in revenue.
On the privacy thing, they wipe the picture of your car after they've found the plate on the list of people who've paid. So as long as you pay, no privacy problem.
It depends how you encrypt it. For example if you use SSL to connect to anonymiser.com for your surfing or hotmail, it's encrypted, but there are no keys you could be compelled to hand over.
for general int code, the 800MHz G4 in my mac is about twice as fast as the 450MHz PII in my old work machine... it only gets faster if you altivec stuff, which no one does (except some clever peeps in apple)
I'm a colour scientist (some of the time), and you're not quite right about colour accuracy.
It depends how you measure it, but for (for example) average delta E on a Macbeth chart, no film can get better than about 10 units because of inter-image effects, whereas even basic digitals get down to just 3-4 units of error.
This so depends. A cheap autofocus 35mm camera is way down at about about 1.5mp. Film resolution depends on ISO, colour vs. b/w, transparency vs negative etc etc but 3k x 2k (ie. 6mp) is typical. No lens can get higher than film resolution:-)
My girlfriend works in the private sector (magazine publisher), I'm public sector (research scientist). Her joke is that her job is to make money, my job is to spend it.
She likes private, because it's fast moving and you have to actually do stuff. I like public, because I have lots of money to spend on toyz, not many deadlines, and a lot of freedom.
The exhibition catalogue includes a whole section on "Ranger Gone Bad 2", so quake is sort-of there. Doom isn't mentioned (AFAIK), which is a bit odd.
Re:Ahead of the other guys? Not really...
on
Apple Drops Mac OS 9
·
· Score: 2, Informative
You're right, everone and their dog has had 2D hardware acceleration for >20 years.
Jobs is talking about using OpenGL to render the desktop. The next windows is supposed to be doing this (though with DirectX), so in that sense, OS X is about 2 years ahead.
AFAIK none of the X GUI toolkits have a working OGL backend yet.
I use SK as the back end of my functional programming language. I added two more: Sl and Sr
Sl a b c => (a c) b
Sr a b c => a (b c)
(I think Turner calls these B and C, corrections pls)
It's pretty easy to see that S/Sl/Sr are the optimal 1-ary combinator set. You hardly need K, and you don't need to generate I during compiles (but it's handy during eval to preserve sharing).
CIECAM97 is probably much more complicated than you can use. It lets you take into account crazy stuff like the colour of the wall behind the screen that the person is using to view your web page.
What people call CIELAB is actually CIELAB76. There is an updated version of the standard called CIELAB94 which is quite a bit better, and also CIELAB2000 (which I've not used yet).
Slashdot Mirror
Thanks for your kind words (blush).
The RPMs should have the rubbersheet stuff inside already (complied into the library).
I've added an extra download for the binary for just the plugin, in case you build from source. Check out the vips-7.8 download page again.
There are some extras in the "extras" bit at the top of the website. nip has only been launched properly for 6 months, so there's no public repository yet.
Oh dear, they should be. I'll check.
Hi, I'm the VIPS maintainer. Thanks for the plug Yarn :)
It has a rubbersheet tool. You make a target image (a black and white chessboard works well), print the target out, take a picture of it (you'll need to pick a particular zoom setting to calibrate), and feed the original (undistored) image and the (distorted) photo into the transformation estimator.
The estimator uses gradient analysis to iteratively discover a 0/1/2nd order transform from the distorted image back to the original. It won't work so well at the image edges (you'll need to crop the corrected image down a bit), but mostly it works pretty well. I use it on 10,000 by 10,000 pixel images without problems. For pincushion/barell distortions it should get you a fixed image to less than a pixel error.
Downside: VIPS is GPL and the transform stuff was done by a friend of mine as part of his PhD. He doesn't want to GPL it, so the transform stuff is distributed as a binary plugin. The windows and suse binaries on the website include it.
You're right, but of course the service is currently restricted to US-resident, Mac-owning, OSX-installed broadband users. When they open the user base out a bit there should be a very significant jump in revenue.
you can click on the "html source" tab at the bottom and just use it as a notepad-alike
How do you titilate an ocelot?
You oscillate it's tit a lot.
On the privacy thing, they wipe the picture of your car after they've found the plate on the list of people who've paid. So as long as you pay, no privacy problem.
http://www.guardian.co.uk/Print/0,3858,4552927,00. html
"as monadic programming is less than 10 years old,"
:) I had monads in my phd thesis in 1989.
15 I think
It depends how you encrypt it. For example if you use SSL to connect to anonymiser.com for your surfing or hotmail, it's encrypted, but there are no keys you could be compelled to hand over.
No typo, I just meant that a PII and a G4 are about the same speed at the same clock in my experience. Fortunately SPEC agree with me :)
for general int code, the 800MHz G4 in my mac is about twice as fast as the 450MHz PII in my old work machine ... it only gets faster if you altivec stuff, which no one does (except some clever peeps in apple)
I'm a colour scientist (some of the time), and you're not quite right about colour accuracy.
It depends how you measure it, but for (for example) average delta E on a Macbeth chart, no film can get better than about 10 units because of inter-image effects, whereas even basic digitals get down to just 3-4 units of error.
This so depends. A cheap autofocus 35mm camera is way down at about about 1.5mp. Film resolution depends on ISO, colour vs. b/w, transparency vs negative etc etc but 3k x 2k (ie. 6mp) is typical. No lens can get higher than film resolution :-)
My girlfriend works in the private sector (magazine publisher), I'm public sector (research scientist). Her joke is that her job is to make money, my job is to spend it.
:-(
She likes private, because it's fast moving and you have to actually do stuff. I like public, because I have lots of money to spend on toyz, not many deadlines, and a lot of freedom.
She's paid 3x more than me though
iPod does mp3 in hardware too :-( if they did software ogg decode, you'd see battery life tumbling
eh? their wp *does* read .doc, read the article
and their site works fine in moz, what makes you think its ie only?
(you're right though, the flash is annoying)
inconclusive? crumbs, are we looking at the same figures?
darwin wins a couple of tests, but generally linux is between 2 and 10 times faster
not that this is such a bad thing, darwin will speed up, it's a young-ish platform that's not been worked on as much as other *nixes
and server performance is not why people buy macs anyway
The exhibition catalogue includes a whole section on "Ranger Gone Bad 2", so quake is sort-of there. Doom isn't mentioned (AFAIK), which is a bit odd.
You're right, everone and their dog has had 2D hardware acceleration for >20 years.
Jobs is talking about using OpenGL to render the desktop. The next windows is supposed to be doing this (though with DirectX), so in that sense, OS X is about 2 years ahead.
AFAIK none of the X GUI toolkits have a working OGL backend yet.
Hi, works fine for me across machines. I think it must be a problem with your X server. Maybe try a newer eXceed? Or one of the free xf86 win servers?
John
I use SK as the back end of my functional programming language. I added two more: Sl and Sr
Sl a b c => (a c) b
Sr a b c => a (b c)
(I think Turner calls these B and C, corrections pls)
It's pretty easy to see that S/Sl/Sr are the optimal 1-ary combinator set. You hardly need K, and you don't need to generate I during compiles (but it's handy during eval to preserve sharing).
John
Hey, that's the post I was going to write! :)
CIECAM97 is probably much more complicated than you can use. It lets you take into account crazy stuff like the colour of the wall behind the screen that the person is using to view your web page.
What people call CIELAB is actually CIELAB76. There is an updated version of the standard called CIELAB94 which is quite a bit better, and also CIELAB2000 (which I've not used yet).
John
Pretty slow in PC terms ... I saw it compared to a PII 266 (sorry, lost the link).
It only gets quick when you start doing asm and working those vector units.
Here's a shot from a few days ago showing reflections off curved surfaces and multiple soft shadows.