Your DREAM SPECS for a Digital Cinema Camera?

[Sam Wells]

I know that the Fujifilm s5 pro still camera has some form of ccd chip which includes sub pixels in a honeycomb formation to provide extra dynamic range.

 

Not sure why you couldn't do this with CMOS.

 

(Anyway Fuji seems to be doing some serious playing with CMOS as I linked previously...)

 

-Sam

[John Brawley]

With film, cinematographers already have to deal with the problem of having too much resolution for some purposes. So they use filters. From this, Tiffen makes a living.

-- J.S.

 

 

John, with greatest respect, i think that's a bit of an over-simplification.

 

Diffusion is more selective than just simply loosing "resolution". Otherwise, why wouldn't you just blur in post ? (well some do but it's not the same) .Doing it optically and in camera has a different effect.

 

A black pro mist is different to even a white pro-mist. A schneider classic soft is different again, as is some stocking stretched over the back of the lens. I've even used a bit of nose grease smudged on the lens in just the right spot. You still need all that resolution to record the *selectively* defused images you are creating.

 

jb

[John Sprung]

John, with greatest respect, i think that's a bit of an over-simplification.

I agree completely. I wasn't trying to characterize all aspects of diffusion filtration. Just responding that we already have sometimes too much resolution, and a solution to the problem of too much resolution -- in fact, a lot of good easy solutions. The hard part is picking one. ;-)

 

 

 

-- J.S.

[Will Earl]

I also have plans to start a camera project. It'll use an 8x8x3 array of SD CCD sensors to achieve a RGB image at a frame size of 5760x4608 and will shoot anything between 24 to 564 fps at that size. It'll have a five terabyte disposible raid for storage of footage and two itunes library. It'll also have voice synthesis system called iDirector which will call out "action","cut" and provide over 1000 different notes for actors and crew and also include a assistant director module (purchased seperately) which will ask how things are moving along and if they could be sped up at all. It'll also shoot stereo by vibrating the lens and sensor back and forth 144 times per second - it can either produce left and right stereo images or a single RGBZ image when shooting in stereo. The camera will also do noise reduction and add grain to the image. It'll also have an air freshner, AM radio, satnav and armrests. The camera will also have a 'tuxedo' adapter which allows you to wear it like a suit.

 

I also need funding for R&D.

[Thomas James]

When I was talking about a fovea camera I was talking about a camera designed on the concepts of the human visual system rather than the brand name fovean sensor. Such a camera would have the pixels arranged in concentric circles and would be space variant meaning that the pixels would be concentrated and would be smaller in the center or fovea of vision and like the human visual system the pixels would progressively get bigger as the it approaches the peripheral areas of vison. The pixels would also be arranged in a hexagonal pattern so as to reduce alliasing. Thus an interesting tile like pattern would emerge in this type of artificial humanoid retina in contradistinction to Jim Jannard's rectilinear grid.

 

Of course this embraces a form of compression as picture resolution is reduced outwardly. However since the angle of vision is so wide there is really no need to invest resolution resources towards the peripheral areas of vision as fine details will not be noticed by most people unless one intentional focuses their attention away from the subjects of interest. Thus such a system will vastly exceed 65mm photography and will be more equivalent to the IMAX experience.

[Ruairi Robinson]

When I was talking about a fovea camera I was talking about a camera designed on the concepts of the human visual system rather than the brand name fovean sensor. Such a camera would have the pixels arranged in concentric circles and would be space variant meaning that the pixels would be concentrated and would be smaller in the center or fovea of vision and like the human visual system the pixels would progressively get bigger as the it approaches the peripheral areas of vison. The pixels would also be arranged in a hexagonal pattern so as to reduce alliasing. Thus an interesting tile like pattern would emerge in this type of artificial humanoid retina in contradistinction to Jim Jannard's rectilinear grid.

 

Of course this embraces a form of compression as picture resolution is reduced outwardly. However since the angle of vision is so wide there is really no need to invest resolution resources towards the peripheral areas of vision as fine details will not be noticed by most people unless one intentional focuses their attention away from the subjects of interest. Thus such a system will vastly exceed 65mm photography and will be more equivalent to the IMAX experience.

 

yeah, but... in movies, sometimes important stuff happens at the side of the frame... so... this idea is rubbish.

 

Yours sincerely,

A fan of widescreen photography.

[Chris Kenny]

When I was talking about a fovea camera I was talking about a camera designed on the concepts of the human visual system rather than the brand name fovean sensor. Such a camera would have the pixels arranged in concentric circles and would be space variant meaning that the pixels would be concentrated and would be smaller in the center or fovea of vision and like the human visual system the pixels would progressively get bigger as the it approaches the peripheral areas of vison. The pixels would also be arranged in a hexagonal pattern so as to reduce alliasing. Thus an interesting tile like pattern would emerge in this type of artificial humanoid retina in contradistinction to Jim Jannard's rectilinear grid.

 

Of course this embraces a form of compression as picture resolution is reduced outwardly. However since the angle of vision is so wide there is really no need to invest resolution resources towards the peripheral areas of vision as fine details will not be noticed by most people unless one intentional focuses their attention away from the subjects of interest. Thus such a system will vastly exceed 65mm photography and will be more equivalent to the IMAX experience.

 

You really don't want to model the human eye too closely in this respect. (Or most others, for that matter.) The human eye's point of sharp focus is unbelievably small. It's literally about one degree. You usually don't realize this because your eyes are constantly scanning and your brain is piecing together a coherent image. That doesn't work for an artificial imaging system. Such a small point of focus would create totally unwatchable images.

 

You'd quite literally be looking at shots where, in a medium close-up of a face, one eye might be in sharp focus, with everything else being blurry. So, for starters, viewers are going to have difficulty recognizing your characters....

 

If you want a more practically sized point of sharp focus that the cinematographer can control, try a Lensbaby. Or create the effect in post.

[Thomas James]

First of all with a foveated android sensor that is capable of a 180 degree field of vision I would think that it would be unwise to concentrate all our resolution resources to the peripheral areas of the human vision which most people would not notice during a theatre presentation. After all we are talking about foveated android sensor which is a compression system and all compression is a compromise in that it throws away picture detail but what compression can we uses that is the least compromising? And after all the human visual system itself also uses compression otherwise our optic nerves would have to be huge to carry all the information.

 

This is not to say that the proportion of the space variancy of the individual pixels would exactly be modeled after the human eye as this may result in either tunnel vision and allowences would have to be made for the fact the audience may prefer to look at other object of interest. Thus with scientific research various tiled patterns of artificial retinas could be developed to see which patterns work the best. And if the retina of the camera could be hot swappable the photographer could have choices of which retinas to use much like he can choose his lenses.

 

As far as the fact that the human visual system uses saccading in order to scan the image and to build a complete picture an artificial machine vision system could also employ robotically controlled saccading in order to build up a picture frame out of individual fields. However unlike interlace scanning these so called fields would compose of complete pictures and a sufficiently high refresh rates would be enabled so as to build the picture up as quickly as possible for a seamless intergration and motion fidelity.

[Ruairi Robinson]

First of all with a foveated android sensor that is capable of a 180 degree field of vision I would think that it would be unwise to concentrate all our resolution resources to the peripheral areas of the human vision which most people would not notice during a theatre presentation. After all we are talking about foveated android sensor which is a compression system and all compression is a compromise in that it throws away picture detail but what compression can we uses that is the least compromising? And after all the human visual system itself also uses compression otherwise our optic nerves would have to be huge to carry all the information.

 

This is not to say that the proportion of the space variancy of the individual pixels would exactly be modeled after the human eye as this may result in either tunnel vision and allowences would have to be made for the fact the audience may prefer to look at other object of interest. Thus with scientific research various tiled patterns of artificial retinas could be developed to see which patterns work the best. And if the retina of the camera could be hot swappable the photographer could have choices of which retinas to use much like he can choose his lenses.

 

As far as the fact that the human visual system uses saccading in order to scan the image and to build a complete picture an artificial machine vision system could also employ robotically controlled saccading in order to build up a picture frame out of individual fields. However unlike interlace scanning these so called fields would compose of complete pictures and a sufficiently high refresh rates would be enabled so as to build the picture up as quickly as possible for a seamless intergration and motion fidelity.

 

Most people DO notice the peripheral areas in theatrical presentation because filmmakers frame stuff in the far corners of the screen and guide the viewers eye there ALL the bloody time.

 

So you are suggesting we trade a system that has equally high resolution on every part of the screen, so the filmmaker can guide the viewers eye wherever he wants at any time, and there is no visual loss of quality anywhere, to a system where the DOP has to swap the bloody sensor out of he wants the left side of the screen to look sharp?

 

You are off yer rocker mate. Never, ever, ever gonna happen, in a million years.

[John Sprung]

Another way to look at this: The real world is full resolution everywhere, and as you walk around in it, you may look with your fovea at just about anything. If you movie screen had only one sharp area, how do you make the whole audience sync up looking at just that one spot? The screen needs full resolution wherever people might look, which is the whole screen. It needs to match the real world.

 

 

 

-- J.S.

[Thomas James]

The real world is not full resolution everywhere. Reality itself is just an observed reality and before reality is observed it exists in a quantum superposition of alternate realities. A pure mathematical model of the human visual system would be one of infinite resolvability. This is because inside the foevea there is sub fovea and inside the sub fovea there is a sub sub fovea and this theoretically continues on into infinity.

[Dan Diaconu M]

Reality itself is just an observed reality and before reality is observed it exists in a quantum superposition of alternate realities.

 

The above seems like a contradiction. Reality IS as it is (whether we see it or not) and it is in full resolution. Very full.

Was kinda sleepeee one mornig and the hard reality of a door (I did not see) wake me up in a brutal manner (dam door and dam reality)

[Thomas James]

Just because one wakes up into a single reality does not mean that other realities do not exist. In fact it has been scientifically proven that not only do other realities exist but that the universe is depends on other realities for it existence. Look at a typical atom for instance. Classical physics based on only one reality models the atom by having electrons orbitong a nucleus. But the only problem with this model is that if this were true the electrons would eventually dive into the nucleus and the universe would be destroyed. What is really happening is that a single electron is stationary but it occupies different positions according to its existence in different universes.

[Dan Diaconu M]

Just because one wakes up into a single reality does not mean that other realities do not exist.

True. Same goes about anything else (God included), but that's not proof by any standards either.

 

My comment was supposed to make some pips laugh (a joke that is) and ease the atm. on this thread.

"other realities" and all "scientific FACTS" were/are subject to further revisions (imho) How many times in our history didn't we yell "I've got it" and half a minute (or century) latter.... I ain't got it? Secondly... all we advocate (about this stuff) is basicaly hear-say. We don't have first hand experience. We only echo what we heard some experts said (unless you have a P/T at cern). Film and video, yes. We can talk as we have first hand experience. I wish you well Tom.

[Sam Wells]

J What is really happening is that a single electron is stationary but it occupies different positions according to its existence in different universes.

 

And you're sure of this.... ?

 

Offer a proof of this and you can start developing your foveal camera with the Nobel Prize money :)

 

The point is we SEE the world as 'resolved' so to speak even if it is not in the quantum sense. Our imaging systems naturally follow. You simply can't invoke quantum mechanics in your defense because there is no good way to visualize quantum states; it's inherently a contradiction. You can invoke metaphors if you like and perhaps even create visual material that works metaphorically but I think that's the limit.

 

-Sam

[Thomas James]

I would not be eligible for a Nobel Prize for an anthropomorphic foveal camera because this is not my invention. The Everett many worlds interpretation of quantum mechanics is not my idea either but could be useful for some type of infinite grand unification theory that combines quantum mechanics with the theory of relativity.

 

I have a proposal for a quantum telescope that can see into the future but the thing I do not like about the Nobel Prize Committee is that the original purpose of the prize money was to offer a good hypothesis and then they would award you the prize money to try and prove to see if your invention really works. Now adays the whole concept of the Nobel Prize is that first you have to prove that your invention works before they give you any prize money which defeats the whole purpose of getting that money in the first place. For the scoffers who say that my invention can never predict the future of the stock market I say that with a conventional telescope we can see into the past but we can never use these telescopes to record any past historical events on earth.

[John Sprung]

.... the thing I do not like about the Nobel Prize Committee is ....

Nobel prizes don't mean squat, they gave one to Arafat. ;-)

 

 

 

-- J.S.

[Thomas James]

Yes concerning Arafat I really do not have a lot of sympathy for terrorists. Yet interestingly before Arafat died he did give an audience to Gandhi who recommended that Arafat use only non violent means to advance his struggle.

[Thomas James]

So back on topic the main criticism of an anthropomorphic visual sensor is that since it concentrates all of the resolution around the central fovea the resultant image will look like "tunnel vision " which will be unwatchable.

The truth is that so far artificial foveated retinas have been constructed in this fashion but only because there was a lack of understanding of the enormous complexities of the human visual system. My proposal is to have an artificial humanoid retina constructed that more faithfully copies the human retina. Although such a retina will concentrate the chroma resolution around the central fovea with the artificial cones suprisingly the artificial rods of this sensor which give only luma resolution will be totally missing in the central fovea and only appear outside the foveas boundaries. Starting outside the fovea the luma resolution then ramps up so that the greatest concentration of resolution is not in the central part of the angle of human vision but rather the middle part of the angle of human vision and then the luma resolution tapers off as it aproaches the peripheral areas of human vision which encompases a 150 degree wide angle view. Such a scheme will work because it is a known fact that the human eye is generally much more sensitive to luma resolution than it is to chroma resolution with the exception of the central part of the fovea that requires more chroma resolution.

Edited July 23, 2008 by Thomas James