"I'll grab a camcorder and interview employees and scientists, broadcast all the public domain educational media I can find, and work with colleges to get special assignments for students to help run the gig."
Good idea, but shouldn't you be constantly calling and emailing NASA scientists, managers and PR-people instead of posting it on/.?
I've registered personally a long time ago with a special email address that I only use for nyt, and I have _never_ received any spam on that email address. They are behaving fine in my book.
The goal of a compression codec is not to keep the pixel values the same, but the goal is to keep the video quality as high as possible given the bit rate restriction. Those are two different things.
A double blind test, although a lot of work to do right, will find all differences that are relevant. Human perception of images and video is the same as audio: lossy. Certain things are less visible or invisible for humans, and for certain other things, humans are unexpectedly sensitive. Researchers have been trying for decades to find algorithms that measure image or video quality, and the current state of the art is pretty much that you can get _a_ measure of quality, but it will have important shortcomings.
Changing the pixels does not mean reducing the video quality. There is no guarantee that the quality goes down when you change the pixels. It's not that simple: Different changes on different pixels have different impact on the quality.
In many ways, it is possible that you can change the pixels of a video and even improve the quality. Many video sources (usually not feature film DVD's though, because they have already been processed quite a bit) contain spatial and temporal noise from the camera. That can be filtered out with filters of varying complexity (a top of the line noise filter algorithm uses more computations than your whole MPEG2/MPEG4 compression algorithm). Besides noise filtering, you often also need to do color correction, contrast, etc. All those algoritms change the pixels and improve the video quality as perceived by a human being. There is another totally different technique to dramatically improve the image quality: superresolution. Superresolution is the technique of combining multiple images into fewer higher resolution images. NASA has been doing that for decades, and I'm sure that when combined with motion estimation it will improve the video quality of many video segments (it will take a lot of CPU cycles though...). The problem with applying superresolution to video is that automatic reliable&precise 3d object recognition and tracking in natural video is not easy, and that our CPU's are orders of magnitude too slow, but that doesn't mean that it doesn't improve the video quality when it changes the pixels. But actually, as an example 'line doubling' which is often used to extract a higher resolution progressive video from an interlaced source, and can be seen as a very primitive form of superresolution.
Given that the DVD MPEG2 bit rate is extremely high compared to the bitrates used here, I think that it doesn't matter much that the original content was MPEG2. Enough that I wouldn't worry about that for this test. I think MPEG2 on DVD's is 5Mbit (correct me if I'm wrong), which comes down to 0.5 bits per pixel, which really a lot of bits for IBP MPEG2, except if you are zooming in at stills of sequences with very extreme image&motion changes (explosions, etc), becasue then the MPEG encoder would have to resort to I-frames, for which 0.5bits per pixel is a little on the low side (depending on the image content (how many macroblocks have irregular surfaces))...
Google is so fast, it can probably (almost?) search its own database to see if it has seen the link already. If the load is too much, then restrict the search to a fraction, such as only once per 25 links in a search branch, or once per second, or maybe just random inspections. Then the robot will loop for a bit, but that's it.
If google is smart, then they'll have robots close to as many servers as possible, preferably at least 1U colocated at each more than insignificant hosting provider, so that crawling bandwidth is cheap and plentiful.
It happens everywhere, but that doesn't mean it doesn't tend to happen more often in countries where it's easier for kids to get their hands on deadly weapons.
The term "World" does not have to mean the entire planet. For many people the world is the kind of place the poster describes. Likewise, in the world of chemistry, everybody agrees that totally pure water is made of H2O, but in the world of religion you will find people who require some additional properties for water to be considered totally pure.
One of the definitions of world in the Webster dictionary of English: "In a more restricted sense, that part of the earth and its concerns which is known to any one, or contemplated by any one; a division of the globe, or of its inhabitants; human affairs as seen from a certain position, or from a given point of view; also, state of existence; scene of life and action; as, the Old World; the New World; the religious world; the Catholic world; the upper world; the future world; the heathen world."
"Your condition that "patents should be limited to actual useful inventions," strikes me as incredibly impractical." "Besides, why should that be the criteria?"
I believe the original poster meant that the patent should be an invention, not just something that many people can think up when they sit down for it.
Practicality is irrelevant. The patent grants a right to the (quoted) 'Inventors'. Yes, the people granted a right under patent law are not the 'patent filers', 'lawywer', 'engineer', 'designer', or whatever, no they chose to use the term inventor for a reason.
The law also has a method of establishing the difference between an incremental trick, and inventions, by the nonobviousness test in the requirements for patents. That means that the patent must be considered non obvious by specialists in the field. Now that is where practicality comes in: The USPTO doesn't check for nonobviousness.
But that doesn't mean it's impossible to do it right.
Given that those motorcycle engines reach up to around 125 horses per liter, and in comparison a 2004 honda civic coupe dx has 115 horses to run on, I stand corrected and replace 'most' with 'a lot of'...
Yeah, especially that she's not just a hot chick on a motorcycle, but she's completely in gear (did you see the 'kawasaki' leather jacket?), and the bike probably has more horsepower than most/. readers' cars...
hihi, it's true. I guess it updates quickly. Maybe it uses 'referrer' to figure out which sites to update first?
Or maybe it knows that all tech stuff on the internet is summarized, ridiculed, flamed about, explained, hated, loved, put into context, taken out of context, etc all at the same time here at slashdot?
Slashdot Mirror
Nah, obviously something like "F150" or "Camry" would sell much better.
"I'll grab a camcorder and interview employees and scientists, broadcast all the public domain educational media I can find, and work with colleges to get special assignments for students to help run the gig."
/.?
Good idea, but shouldn't you be constantly calling and emailing NASA scientists, managers and PR-people instead of posting it on
I've registered personally a long time ago with a special email address that I only use for nyt, and I have _never_ received any spam on that email address. They are behaving fine in my book.
The goal of a compression codec is not to keep the pixel values the same, but the goal is to keep the video quality as high as possible given the bit rate restriction. Those are two different things.
A double blind test, although a lot of work to do right, will find all differences that are relevant. Human perception of images and video is the same as audio: lossy. Certain things are less visible or invisible for humans, and for certain other things, humans are unexpectedly sensitive. Researchers have been trying for decades to find algorithms that measure image or video quality, and the current state of the art is pretty much that you can get _a_ measure of quality, but it will have important shortcomings.
Changing the pixels does not mean reducing the video quality. There is no guarantee that the quality goes down when you change the pixels. It's not that simple: Different changes on different pixels have different impact on the quality.
In many ways, it is possible that you can change the pixels of a video and even improve the quality. Many video sources (usually not feature film DVD's though, because they have already been processed quite a bit) contain spatial and temporal noise from the camera. That can be filtered out with filters of varying complexity (a top of the line noise filter algorithm uses more computations than your whole MPEG2/MPEG4 compression algorithm). Besides noise filtering, you often also need to do color correction, contrast, etc. All those algoritms change the pixels and improve the video quality as perceived by a human being. There is another totally different technique to dramatically improve the image quality: superresolution. Superresolution is the technique of combining multiple images into fewer higher resolution images. NASA has been doing that for decades, and I'm sure that when combined with motion estimation it will improve the video quality of many video segments (it will take a lot of CPU cycles though...). The problem with applying superresolution to video is that automatic reliable&precise 3d object recognition and tracking in natural video is not easy, and that our CPU's are orders of magnitude too slow, but that doesn't mean that it doesn't improve the video quality when it changes the pixels. But actually, as an example 'line doubling' which is often used to extract a higher resolution progressive video from an interlaced source, and can be seen as a very primitive form of superresolution.
He probably means umlwin32
Given that the DVD MPEG2 bit rate is extremely high compared to the bitrates used here, I think that it doesn't matter much that the original content was MPEG2. Enough that I wouldn't worry about that for this test. I think MPEG2 on DVD's is 5Mbit (correct me if I'm wrong), which comes down to 0.5 bits per pixel, which really a lot of bits for IBP MPEG2, except if you are zooming in at stills of sequences with very extreme image&motion changes (explosions, etc), becasue then the MPEG encoder would have to resort to I-frames, for which 0.5bits per pixel is a little on the low side (depending on the image content (how many macroblocks have irregular surfaces))...
The problem is that SNR, or even PSNR is not really a good measure. The only good measure is with a double-blind test with a lot of people...
"I wonder how many e-mails a zombie Pocket PC can crank out before the the user sees a $10,000 for bandwidth usage?"
Pocket PC? The battery will be empty before message number six...
Google is so fast, it can probably (almost?) search its own database to see if it has seen the link already. If the load is too much, then restrict the search to a fraction, such as only once per 25 links in a search branch, or once per second, or maybe just random inspections. Then the robot will loop for a bit, but that's it.
If google is smart, then they'll have robots close to as many servers as possible, preferably at least 1U colocated at each more than insignificant hosting provider, so that crawling bandwidth is cheap and plentiful.
CNN has the scoop.
"but it means I would be extremely surprised to hear on the news that 2 Norwegian kids machinegunned their classmates in the school."
World-wide, there are many deadly school murders by students, including in low-crime countries such as Japan.
How about when an eight year old shoots at a school in Germany? Or an expelled student killing 18 in a school in Germany?
It happens everywhere, but that doesn't mean it doesn't tend to happen more often in countries where it's easier for kids to get their hands on deadly weapons.
"USA != world."
That what you said, not what the poster said.
The term "World" does not have to mean the entire planet. For many people the world is the kind of place the poster describes. Likewise, in the world of chemistry, everybody agrees that totally pure water is made of H2O, but in the world of religion you will find people who require some additional properties for water to be considered totally pure.
One of the definitions of world in the Webster dictionary of English: "In a more restricted sense, that part of the earth and its concerns which is known to any one, or contemplated by any one; a division of the globe, or of its inhabitants; human affairs as seen from a certain position, or from a given point of view; also, state of existence; scene of life and action; as, the Old World; the New World; the religious world; the Catholic world; the upper world; the future world; the heathen world."
"Your condition that "patents should be limited to actual useful inventions," strikes me as incredibly impractical." "Besides, why should that be the criteria?"
I believe the original poster meant that the patent should be an invention, not just something that many people can think up when they sit down for it.
Practicality is irrelevant. The patent grants a right to the (quoted) 'Inventors'. Yes, the people granted a right under patent law are not the 'patent filers', 'lawywer', 'engineer', 'designer', or whatever, no they chose to use the term inventor for a reason.
The law also has a method of establishing the difference between an incremental trick, and inventions, by the nonobviousness test in the requirements for patents. That means that the patent must be considered non obvious by specialists in the field. Now that is where practicality comes in: The USPTO doesn't check for nonobviousness.
But that doesn't mean it's impossible to do it right.
What is new? We're not talking about a minivan.
Motorcycles have historically been associated with a combination of danger, power, speed, and substance abuse.
Bikes are not marketed to soccermom.
Given that those motorcycle engines reach up to around 125 horses per liter, and in comparison a 2004 honda civic coupe dx has 115 horses to run on, I stand corrected and replace 'most' with 'a lot of'...
Looks like this guy took that tour last year...
"good luck convincing anyone to go to ground 0 and clean it up (rather than forcing them to do it at gunpoint.)"
If anything, they're not going to just redo the on-site concrete-pooring again, it will be more some sort of prefab rapid construction.
For example: babelfish translated from 3sat.de: Some smart people have been working on that.
thank you, but I come in peace. Take me to your lizard.
That was exactly her and her nuclear-effects-researcher dad's point...
The most dangerous part of driving there, in their opinion, is not the radiation, but the danger resulting from the high speeds on the motorcycle.
"Down quickly; any mirrors out there?"
/. the wayback machine?)
Yes. The WayBack Machine has at least some of it.
(can we
Well, comparing the few before and after pictures on the site, I'd say that it used to be a lot greener there before the accident...
Yeah, especially that she's not just a hot chick on a motorcycle, but she's completely in gear (did you see the 'kawasaki' leather jacket?), and the bike probably has more horsepower than most /. readers' cars...
If Kawasaki picks up on this, they could be in for a hugely successful worldwide marketing campaign around her.
hihi, it's true. I guess it updates quickly. Maybe it uses 'referrer' to figure out which sites to update first?
Or maybe it knows that all tech stuff on the internet is summarized, ridiculed, flamed about, explained, hated, loved, put into context, taken out of context, etc all at the same time here at slashdot?
O.K. Now I have to click on that link at least once!
I used this: XFree86 "4.4"
and that search also returned xfree86.org as the first link.
So to me it does not look like "some porn sites plastered the term "XFree86" all over themselves", but more like a suspiciously blocked search query.