General Demosaicing Questions

[Dj Joofa]

I watched it so long ago, enjoyed the MTF stuff but was very put of by the presentation on Bayer patterns. If you look back in time to my comments I made, I'm sure they'd explain my objections to the video.

The mistake about Nyquist is that sampling theorem doesn't talk about resolution, but about reconstruction of an input signal with respect to the appearance of aliasing or not, in that if you bandwidth limit according to the theorem you can completely reconstruct the sampled waveform, and of course, aliasing would preclude accurate reconstruction.

Hi Graeme,

I know that topics, such as sampling, are close to your heart. But, I think you are disproportinately criticizing some of the presentations. I mentioned before also that one should have an overall appreciation of the effort. Because, if one is to nitpick then any small slip by somebody is an invitation. Because, IMHO, your own comments on sampling theorem are open to criticism unless you realize that the reconstruction as envisioned in the sampling theorem is not directly applicable to image sensors, unless properly accounted for. But, the idea is not to nitpick on a few technical errors, but to realize that such presentations are for public education.

Sincerely,

Joofa

 

[Jannard]

but to realize that such presentations are for public education.

Sincerely,

Joofa

And why do we want to educate the public? If nothing is to be done with the info... there is no point sharing. If something is to be done with the info... it is competition.

Jim

 

[Peter Moretti]

Jim,

I realize that you have to protect your intellectual property, and only you know how much Red's policy of openness might have cost your company.

But please realize that there is a very real upside in the dialogs that have taken place. People who understand things like Bayer patterns and wavelet compression are better equipped to defend using Red as a choice of camera.

If it weren't for Graeme's posts, I would never have been able to say to my friend that he should use the Red One b/c the F35's "true 4:4:4" comes with the price of chroma aliasing.

As smart as people like Graeme and Stuart are, I don't think they are going to teach the engineers at Sony or Canon much.

It's the fun, skunkworks environment that you've created at Red that makes the company a success, not any idea about sampling theory that no one else has thought of.

Even the concept of the Red One, a RAW camera for moving images, many people thought of it. But Red took that idea and ran with it and made it a reality. It wasn't the idea. It was YOU, and those around you, that made RAW a reality.

So I'm sure you'll protect what you need to protect, but the secret sauce is staring at you right in the mirror.

 

[Christopher Probst]

Yea Jim, I too battle the idea of how much "proprietary" info to share. Is there such a thing as a trade secret? Sure filmmaking is a craft handed down and taught to each new generation, but it's also a business. WIth competition. I could easily list out all of my various methods and tricks of doing what I do, what I feel gives me an edge over the next guy doing similar jobs to me. But I don't.

However, on some level I do also try to give something back. Most of the thrust of American Cinematographer is weighted toward the newer generation of filmmakers. Now, this is my opinion, but I believe the magazine is not intended for just a hundred or so ASC members to read, but rather the 10s of thousands aspiring to learn their craft in the tradition of ASC cameramen. I didn't write my special lab processing article [AC Nov '98] to teach Janusz Kaminski about ENR processing. I wrote it for those that have yet to use it and want to learn more about what it exactly is.

So I do understand your meaning Jim. And all evidence to the contrary, you ARE educating the public on these forums daily, like it or not!

 

[Jannard]

Yea Jim, I too battle the idea of how much "proprietary" info to share. Is there such a thing as a trade secret? Sure filmmaking is a craft handed down and taught to each new generation, but it's also a business. WIth competition. I could easily list out all of my various methods and tricks of doing what I do, what I feel gives me an edge over the next guy doing similar jobs to me. But I don't.

However, on some level I do also try to give something back. Most of the thrust of American Cinematographer is weighted toward the newer generation of filmmakers. Now, this is my opinion, but I believe the magazine is not intended for just a hundred or so ASC members to read, but rather the 10s of thousands aspiring to learn their craft in the tradition of ASC cameramen. I didn't write my special lab processing article [AC Nov '98] to teach Janusz Kaminski about ENR processing. I wrote it for those that have yet to use it and want to learn more about what it exactly is.

So I do understand your meaning Jim. And all evidence to the contrary, you ARE educating the public on these forums daily, like it or not!

There is room for lots of DPs... but only a few camera companies. RED will be alone in manufacturing in the US, at some expense. We are OK with that because we want to put people to work here. But putting people to work is also a responsibility. We need to stay in business to make sure the paychecks last a long time. RED gives up lots of "muscle" to the Japanese as far as production capabilities and efficiencies. About all we have is ingenuity to keep us in front. Giving that away puts our team at risk. I have seen this for many years starting at Oakley.

We don't mind giving out certain info and Graeme does a good job metering it out. Where we draw the line is answering every question, especially when it gets close to the family jewels. It is up to use where to draw the line, not posters or reporters.

I have a lot of respect for you. I just hope you will respect our need to keep some things close to the vest.

Jim

 

[Jay Gannon]

As smart as people like Graeme and Stuart are, I don't think they are going to teach the engineers at Sony or Canon much.

I don't know about that, generally the people who are so far above their peers in their field work for small skunkworks like RED, and not for Sony or Canon.

In particular algorithms and other such IP are by in large the product of a single and amazing brain.. who I believe in this case in Graeme with RED, he may have built upon a solid foundation, but I think credit needs to be given that he is probably among a group of less than 5 people who operate at the level he does when it comes to his profession.

So I think protecting the IP on people like Graeme and Stewart is very important and should be the one thing that sets RED apart from everything else, with Epic that will probably extend to your electrical and electronic engineers who have done some crazy things with shrinking the package, but at the absolute heart of RED is REDCODE which is Graemes baby as far as I know. Without that a RED One would need to be strapped to a massive recorder and would lose its price point its flexibility and everything else. Not even to speak of the Epics size/flexibility price point.

So while RED is a very very open company, I think that we should be not quite grateful but very privileged that Graeme, Stewart, Deanan et al. are not locked up in a box with all of their communications monitored for fear they will reveal corporate IP.

The mere fact that we get to talk to any lead engineer on a product that we use is pretty revolutionary. Because many of his peers in other companies are to be honest and I'm sure he/jim can probably attest to that fact.

 

[Peter Moretti]

Jay,

One-on-one, I think think Graeme or Stuart wax a lot of the guys at Sony and Canon. But those companies have teams of engineers, and they design and manufacture their own chips. Their engineering depth and breadth is probably far beyond what we realize.

What Red can do is nimbly react w/o layers of management to deal with and legacy products to protect.

Where Red is hamstrung is manufacturing, as Jim alluded to. I once floated the idea that Red and Nikon should team-up (got reamed for it by various posters). And who knows how that type of relationship could actually work? I believe Nikon also doesn't make its own chips, but they certainly do make cameras and lenses and they direly need help in video, b/c Canon is kicking their a$$. I DO think it's cool that Red is trying to bring camera manufacturing back to the U.S.! (But maybe Nikon could make models like the fixed Scarlet, which I believe Jim said will not be made here?) But it's probably just a pipe dream b/c I happen to be a Nikon fan.

But I do believe that manufacturing is going to continue to be a big issue for Red. And I trust that Jim and everyone will know what to tell and what to keep close to the vest.

In the final analysis, I think Red's culture is its biggest advantage.

 

[Jay Gannon]

Until today I would have tended to agree but Jims post:

A few weeks ago we hired the person responsible for setting up all of Oakley's production. I worked with him for 15 years. He is the best.

RED is no longer a development company. Results should become apparent soon.

Jim

Would lead me believe that there is a shift in ethos at RED to a longer development cycle with the Epic and a larger emphasis on production and developing the product base.

If Jim is starting to hire hardcore production staff then that can only be a good thing for RED (Sorry Jim for talking about you in the third person btw) and would indicate to me the decision has been made to proceed with an expansion of in-house production facilities and not to partner with other company.

 

[David Rasberry]

Even with their size, I doubt Sony Et Al have significantly larger teams of engineers devoted to development of a single product, especially high end video cameras. They produce a broad range of products of all kinds.

 

[Dj Joofa]

And why do we want to educate the public? If nothing is to be done with the info... there is no point sharing. If something is to be done with the info... it is competition.

Jim

Dear Mr. Jannard,

I understand your concern, however, I respectfully beg to disagree with your viewpoint. Nobody is asking Red to share internal information. You can have a look at the Panavision videos and decide if they leaked any secret sauce. But, the presentations were still very good for public education.

I think at this stage Red should feel as a more mature company that can talk about technical stuff without letting the cat out of the bag.

Sincerely,

Joofa

 

[David Rasberry]

The best commentary and education about how RED fares against the other guys is Graeme's posted zone charts and Red's sample reels.:thumbup1:

 

[Bill Anderson]

Guys, what's so difficult to understand about a camera manufacturer keeping a lid on certain procedures? And how does that necessity compare with the spirit of sharing our technical experience with other artists? In my mind they are two very different things. If sharing a proprietary method meant that you and your friends might never work again, would you be so generous, or nonchalant about the issue?


And what is all this about how much resolution, (or anything else for that matter) we, as artists, might "need"? Because a handful of image-makers are using "old glass", or others are happy enough with a "2K" finish, that is hardly grounds for setting standards for others to live by: No matter how brilliant, no filmmaker to date has managed to make technological advancement superfluous. So, regardless of who you are or what you have done, your work will never be understood as an absolute, something that will, Godlike, preside over every future artist's endeavors.

 

[Eric Lange]

Almost by accident… having a better understanding of things (only just this morning) I am leaning much more to the idea that divulging directly the inner workings of certain processes in a dumbed down conceptual form frankly does not in this case make for a good marketing tool. I think (now) Jim is 100% correct that there is absolutely no advantage. In some cases it can be useful, in this case it is clearly not. Even if the facts were presented in some accessible non-damaging form (in this case); people and organizations unfortunately have a habit of taking that information and completely twisting and skewing everything against you as a negative marketing tool. The competing marketing guys don’t care about math, they instead look for catchy images and phrases that stick in the mind that are consequently very hard to undo.


Hats off to RED as they are not given to breaking things out into various BS marketing handles. From a marketing perspective you can invent all kinds of new boxes and handles layered on top of the actual processes. So you could say, … actually I’m not even going to say what you could say…


As usual half the time I make a complete U-turn (funny how that happens).[(I guess that’s the idea of constructive discussion that one is open to having one’s mind changed.)]


I think the other thing is that people vastly underestimate the collective “pig headedness” of various large corporations that are very used to doing “their thing”, “their way”. If you mailed all the specs, relevant IP and physical units to the various respective corporations they would perhaps glance at the material and say ”so what?”, even if it was the most brilliant solution on the planet.


It is possible to reverse engineer all kind of things, that is why unfortunately we have to rely on a faulty patent system. Most invented systems are not beyond the wit of man, if the will is there to crack something it can be done. This basically returns to the idea that Red is the best in the world at being Red and delivering RED cameras and challenging themselves to be out in front no matter what. If you want to jump on the RED bus, you can! SONY “PANNY” ARRI and all the others are not going to make RED (ish) or RED like cameras. It’s not in their DNA and they never will.

Cheers,

Eric

 

[Graeme Nattress]

Eric, that's why I prefer the discussion format we have here, rather than writing static articles and fielding the inevitable emails asking for clarification or questions. I still get emails about articles I wrote 6 or more years ago...

Graeme

 

[Eric Lange]

I have to admit It’s a pretty unique, and strange record that we leave behind. [I think in some instances I should do some spring cleaning of older posts (for my own good).]

I think now that I have a better sense of things “nice” is a good adjective. What you do, does indeed make and make for “nicer” images.

Cheers,

Eric

 

[Bill Anderson]

I like the quality of the images, and where this all seems to be heading at the moment. Call me a romantic, but I would love to see what the likes of David Mullen and C Probst would do with a large sensor. And I mean LARGE. Many of us might squander the additional info on frivolous ends, but there will be a few who, informed by the (indispensable) weight of the historical narrative, will show us old truths in a new light.

I'm not sure we're even remotely aware of the possibilities.

 

[Peter Moretti]

Guys, what's so difficult to understand about a camera manufacturer keeping a lid on certain procedures? And how does that necessity compare with the spirit of sharing our technical experience with other artists? In my mind they are two very different things. If sharing a proprietary method meant that you and your friends might never work again, would you be so generous, or nonchalant about the issue?

...

Of course Red should NOT share proprietary info. And only Red can decide what's classified and what isn't. And I say err on the side of caution.

But let's look at the most successful competitor to date: the Arri Alexa. That camera has very little in common with the Red One and even less with the Epic. It's a different approach to solving a problem. It actually avoids some of Red's advantages and tries to turn them against Red, namely workflow and pixel count.

Look at Sony's SRW-9000PL. It's essentially an F35 with an SR recorder. No Red technology there. Look at the F3: large pixel sensor, no RAW, baked image. How could Sony have designed that camera from reading Reduser? (Unless they decided to do the opposite.)

And look at their upcoming 4K camera; Q67 is described as essentially being bullshit here b/c a Bayer sensor with the same pixel count could resolve even more. (BTW, how do I know that? Only b/c I've read Graeme's posts on the matter. So if someone starts yapping about Q67, I have a retort.) Sony probably made Q67 just b/c they can, i.e. they manufacture their own chips and wanted to stick some new design into the market to show-off that capability. Or maybe they do have a brilliant, low power solution. But again, where is the Red design influence? I don't see it.

The AF100 is based on Panny's GH-2 and is AVCHD. Canon's XF's use 1/3" inch chips and MPEG-2 compression. Canon's DSLR's are based on still camera technology and were essentially a happy accident for the company. Canon's rumored merging of video and DSLR is supposed to be just that, a mix between the XF's and the D's. How does any of this borrow technical knowhow from Red?

I really see no evidence of makers using information gleaned from these boards and utilizing it in present or announced cameras. Don't forget, the other guys are wicked smart too and have egos as well. They probably want to solve the problem their way, and their way may turn out to be very good too.

Does that mean they don't snoop around here? I'm sure they must. But I don't think they actually employ anything they've read here in their own designs. They probably leave thinking one of two things: "What a cool place to work," or "Those guys are CRAZY."

What I do know is that the education here has enabled people to defend the Red One on its technical merits and explain to certain degree why it excels.

But really, all we know about the Red is that it's a S35 sized Bayer CMOS sensor with 13.8 MP that records RAW in the .R3D format. The vast majority of technical discussions revolve around describing in detail what all that means. And operators, owners and DP's need or needed that education. But the folks at Sony, Panavision, Canon, they know it already.

It's like the Wikileaks leaks. Does anyone REALLY think that the other countries didn't know what was leaked? That Saudi Arabia fears Iran more than Israel, that we talk shit behind our allies' backs? That the Middle East states all do that to each other? There are a lot of books out there on Amazon by former and current policy people: Richard Haass, Robert Gates, Robert McNamara, Donald Rumsfeld, that explain exactly that. To use a Rumsfeld-ism (sorry, I'm reading his book right now) these are "known knowns."

So YES, of course Red should protect its trade secrets. But I honestly don't see an indication of anything technical said here being co-opted by the competition.

 

[Anderson Boyd]

There is room for lots of DPs... but only a few camera companies. RED will be alone in manufacturing in the US, at some expense. We are OK with that because we want to put people to work here. But putting people to work is also a responsibility. We need to stay in business to make sure the paychecks last a long time. RED gives up lots of "muscle" to the Japanese as far as production capabilities and efficiencies. About all we have is ingenuity to keep us in front. Giving that away puts our team at risk. I have seen this for many years starting at Oakley.

We don't mind giving out certain info and Graeme does a good job metering it out. Where we draw the line is answering every question, especially when it gets close to the family jewels. It is up to use where to draw the line, not posters or reporters.

I have a lot of respect for you. I just hope you will respect our need to keep some things close to the vest.

Jim

The amount of transparency and open dialogue between creators and users is already monumentally generous. The old adage applies to everything, give someone an inch, they'll want a mile...and I think that's a testament to the excitement and wonder you guys inspire. I've purposely held off in buying anything more than a prosumer camera for years because RED's dedication to its user base is as strong as its dedication to innovation, and I've waited to jump in line for Epic X or S.

My meager post count, 15 I think it is now, doesn't reflect the years I've been reading this site, and each day's reading just bolsters confidence in the company which was strong from the start. It's been a wild ride watching the technology grow as the company does, and I fully believe RED won't just continue to change cinema and television, but every single recorded medium out there (medical, security, etc.)

All of this plants the seed of loyalty in (most) all of us that read this site, and I have no doubt a RED user of today, is the RED user of tomorrow, and the next day, and the next week, and the next year, and the next decade. So tell us as little or as much as you want, you won't hear me complaining. Keep making the world a better place gentleman.

 

[Eric Lange]

Well this is not exactly a De-mosaicing question, but rather an Image compression question so it has nothing to do with REDCODE or De-bayer exactly, more of a “Brain-fart”.


What happened to Fractal image compression? In the late eighties and early 90s it held the promise of really terrific up-scaling, i.e. it “knew” how to make up plausible detail, and used plausible detail to compress an image. It seems to have vanished almost without trace? What happened?

www.ams.org/notices/199606/barnsley.pdf


Also on that note, I’m wondering if a separate general IMAGE SCIENCE or IMAGING SCIENCE section should be set up on REDUSERNET. There seem to be so many generalized sensor and image science questions and threads, I wonder if they should have their own forum/section? So folks can talk about , De-mosaicing questions, alternate compression schemes, Modulation transfer functions, point spread functions, LUTs, color science etc. etc. or specific signal processing or electronic issues CCD vs CMOS, how does it work and so on and so forth or even talk about more abstract optical issues and aspects of optical design. The 3d section on Reduser was set up because S-3d was, and is still is an important application area, but it always seemed to chop threads up when the “3d-ers” would sort of dog-pile a discussion that also had a predominantly 2d cinematic practical thrust to it.


Perhaps an IMAGING SCIENCE section might gather up a lot of disparate threads, people and queries? It might actually form the basis of a useful and very current resource?

I’m always endlessly impressed by the number of professional and amateur engineers, scientists and designers from diverse disciplines that always have something interesting, thought provoking or funny to contribute on REDUSER.

Just a thought.


Cheers,


Eric

 

[David Rasberry]

What happened to Fractal image compression? In the late eighties and early 90s it held the promise of really terrific up-scaling, i.e. it “knew” how to make up plausible detail, and used plausible detail to compress an image. It seems to have vanished almost without trace? What happened?



Eric

onONe Software sells a fractals based program for enlarging digital photos. The opposite of compression really. Used to be called Genuine Fractals, now it is called Perfect Resize. So in that sense fractals are alive and well.