The videos are stored twice, once in 2D, once in stereoscopic 3D, so you're dealing with 140 minutes of video.
This direct grab is littered with compression artifacts. Let's count those blocks.
Digital Foundry Article Technical Discussion Archive [2011]
- Thread starter user542745831
- Start date
-
- Tags
- digital foundry
- Status
This direct grab is littered with compression artifacts. Let's count those blocks.
Hmm I wonder if that's true, maybe binks cpu load is just far less than other options. Is it also possible that bink just works better at really low bitrates compared to h264 and vc-1? Since their videos are meant to hide loading then keeping the video bitrate low is important. I've read that h264 doesn't work as well at low bitrates as say vc-1, and my own video encoding work pretty much confirms that I can get better quality video at low bitrates with vc-1 than h264. Of course Sony would never use vc-1, but maybe that's why they choose bink over h264 since it better suits their target bitrate?
http://forum.beyond3d.com/showpost.php?p=1525909&postcount=1406
One of the 3D modes that displays can detect is a packed buffer: they pack the individual view into the left and right halves of a 1280x720 or 1920x1080 (in Crysis 2's case) buffer. Of course, this means you'll be losing a lot of information in the former due to the downscale and re-upscale, assuming you're starting with full (or close-to) frame renders (2D+depth or the fake parallax that the Crysis 2 beta employed). Were they to render both views at half-res in the first place, it wouldn't be an issue (proper 3D parallax).
It's rather curious that 360 hardware just simply doesn't output the 1280x1470, which is well within the HDMI cable bandwidth. Grandmaster actually covers this a bit here:
http://www.eurogamer.net/articles/digitalfoundry-xbox360-3d-ready-article
Digital Foundry said:
ShadowRunner
Veteran
Its the other way around. They had disk space to burn so using bink, which requires higher bitrate compared to more rescource hungry algorithms in order to get a decent look, wasnt an issue. The more rescource hungry the algorithm the better it tends to look at lower bitrates. It is compressed more so is more intensive task to uncompress.
Last edited by a moderator:
Oh ok thanks for the information. All I need to know you have provided
ihamoitc2005
Veteran
You must research this a bit more, my friend
Actually it does. It does not interpret what it sees as double number of objects but it perceives twice qualitative information (actually more through more accurate estimation of unrendered or inbetween detail) about the object. This has been shown many times in tests where subjects can percieve vastly more information about a scene in 3D than in 2D in less time. Understanding depth and object placement relative to another at a glance is just one of many benefits of stereoscopic vision that we have evolved.
You are right that the brain mixes information. That is one reason why stereo 3D (unlike 2D, each eye sees a different image) can be lower res but give much more information.
This is possible because the brain sees the both information and tries to fill in the gaps so to speak. This is the value of true stereoscopic 3d vs fake image-warping 3D. In true stereoscopic 3D each eye sees completely different color values across objects. It aligns the geometry and uses the different color information due to subtle difference in lighting angle, shadow darkness, etc to create more detailed understanding of information than is possible with simple 2D image.
Pixels have color (or brightness) values. The brain aligns the geometry and extapolates qualitative information based on things such as subtle differences in color values on each point of an object and fills in missing or inbetween details based on the two values resulting from different angle of each eye relative to position x on object and light source. It cannot do this in 2D. In truth, stereoscopic 3D gives vastly more, not just twice information as 2D.
Actually it does. It does not interpret what it sees as double number of objects but it perceives twice qualitative information (actually more through more accurate estimation of unrendered or inbetween detail) about the object. This has been shown many times in tests where subjects can percieve vastly more information about a scene in 3D than in 2D in less time. Understanding depth and object placement relative to another at a glance is just one of many benefits of stereoscopic vision that we have evolved.
You are right that the brain mixes information. That is one reason why stereo 3D (unlike 2D, each eye sees a different image) can be lower res but give much more information.
This is possible because the brain sees the both information and tries to fill in the gaps so to speak. This is the value of true stereoscopic 3d vs fake image-warping 3D. In true stereoscopic 3D each eye sees completely different color values across objects. It aligns the geometry and uses the different color information due to subtle difference in lighting angle, shadow darkness, etc to create more detailed understanding of information than is possible with simple 2D image.
Pixels have color (or brightness) values. The brain aligns the geometry and extapolates qualitative information based on things such as subtle differences in color values on each point of an object and fills in missing or inbetween details based on the two values resulting from different angle of each eye relative to position x on object and light source. It cannot do this in 2D. In truth, stereoscopic 3D gives vastly more, not just twice information as 2D.
Last edited by a moderator:
Exactly how are you reconstructing the texture information... Your imagination doesn't really count if the information isn't there to begin with. Just look at the 3D vs 2D screenshot and you're simply missing a ton of resolution from the lower res mipmaps. No amount of imaginary interpolation is going to give you back the information. That's just silly. You're certainly not going to interpolate the specular map that is entirely missing on the wall in the 3D shot.
L. Scofield
Veteran
The way he explains would be accurate if the game used interlaced frames, but I don't think that's the case.
Taking your stereo sound track example. The way you suggested in your example is:
8bit mono sound vs 8bit stereo sound (assuming frequency is the same).
The right way to do this comparison is:
8bit mono sound vs 4bit stereo sound (assuming frequency is the same).
K, should have used 16bit vs 8bit, but I was trying to make a point.
Voodoo2 is also sooo 1998.
ihamoitc2005
Veteran
Taking your stereo sound track example. The way you suggested in your example is:
8bit mono sound vs 8bit stereo sound (assuming frequency is the same).
The right way to do this comparison is:
8bit mono sound vs 4bit stereo sound (assuming frequency is the same).
K, should have used 16bit vs 8bit, but I was trying to make a point.
Actually, this is incorrect. Bit depth determines dynamic range. This is not analogous to screen resolution because 3D and 2D video have same dynamic range. Analogous to screen resolution is sampling rate. Sampling rate is frequency information. This is something like resolution.
When you listen to stereo audio of the same total bit-rate as mono audio, each ear gets different information and the brain combines to create a total perception that is much more than sum of it's parts.
Same way, when you watch s3D video, each eye gets unique image that the brain combines to get a total perception that is much more than the sum of it's parts.
This is because stereo information taps into our brains ability to infer additional information from what is given in a way that mono cannot.
ihamoitc2005
Veteran
Interlacing
Actually, the brain automatically interlaces different images for each eye by aligning geometry in mental map and then combing color/brightness information from each image. The farther away the object the less this effect has (because distant objects have less difference between left and right image) but for nearby objects it is a very great benefit. This is why we evolve stereo vision and stereo hearing.
Our eye does this every day with our actual eyes looking at real objects. Our eyes are not interlaced. Same with our ears.
Actually, the brain automatically interlaces different images for each eye by aligning geometry in mental map and then combing color/brightness information from each image. The farther away the object the less this effect has (because distant objects have less difference between left and right image) but for nearby objects it is a very great benefit. This is why we evolve stereo vision and stereo hearing.
Our eye does this every day with our actual eyes looking at real objects. Our eyes are not interlaced. Same with our ears.
L. Scofield
Veteran
It's not really interlacing. In this case, since both frames are progressive you're still going to see blurry IQ and double sized jaggies, just at double the frequency of 2D.
Yes… I mentioned we need to take care of 3D videos in my previous post. There is still a factor of 4-5 between the KZ3 compressed video storage and C2's intro segment. Assuming the 3D videos take double the storage, the KZ3 videos should still have better bitrate on the average (Yes, the original question was "I hope KZ3 has better bitrate").
This direct grab is littered with compression artifacts. Let's count those blocks.
Could be worse if they use lower bitrate.
EDIT:
H.264 and VC-1 have multiple profiles, some are designed for low bitrate environment. If you use the right profiles, they should be comparable. I thought VC-1 is no longer actively pushed by MS because the implementors have to pay MPEG-LA patent fees (VC-1 violated MPEG4 patents). Given a choice, Sony may still go for H.264 because their implementations are probably more well used and optimized.
EDIT 2:
Would be nice if we know what the bitrate is.
ihamoitc2005
Veteran
No. I suggest research audio and visual perception and neuroscience.
No. The brain takes 2 full progressive fields and combines different left and right information for each point of geometry. This is not difficult to understand if you stop thinking that the brain operates the way you render images. It will help you to receive a gist of how the brain interlaces the different images if you think of lenticular 3D system with 2 full progressive scan inputs (reminder: actual brain processing is much more sophisticated).
Your brain does this every day at every thing in the real world that you look and hear. Yes, your brain is much smarter than you think.
No. The brain takes 2 full progressive fields and combines different left and right information for each point of geometry. This is not difficult to understand if you stop thinking that the brain operates the way you render images. It will help you to receive a gist of how the brain interlaces the different images if you think of lenticular 3D system with 2 full progressive scan inputs (reminder: actual brain processing is much more sophisticated).
Your brain does this every day at every thing in the real world that you look and hear. Yes, your brain is much smarter than you think.
Last edited by a moderator:
L. Scofield
Veteran
Your mistake is thinking that our brain perceives 3D displays the same way it perceives the natural world. It doesn't. Otherwise we wouldn't have to simulate effects like motion blur.
The brain will interpret the data as 3D, but as blurry 3D
I concur. The information the brain can interpolate/extrapolate is a different type of information to image resolution and texture resolution. It can't necessarily magically ignore big chunky pixels or chunky texels, although a left and right sample at the same image coordinate will result in a data perception different to a single sample. Very little we see is actually 'seen'. That image that your brain presents to your conscious mind is almost entirely an evaluative construct, with a titchy splodge of the real world light captured at the point of focus. A 3D image will be great for filling in general information of the scene and in providing an understanding of the world, but cannot overcome the fundamental resolving power of the eye making out pixelation or trying to fill in missing information. in fact brains tend to do the opposite, decreasing information from a scene to get to the important bits, so cannot be expected to look at a texture sampled at two adjacent texel rendered in the same pixel space and interpret missing detail.
- Status
Similar threads
- Locked
- Deleted member 11852
- Gaming Technology
- Locked