sony

[I Bloom]

For one target pixel, take the value for G from the green diamond.
i.e. (g = G);

For the same target pixel, take the top and bottom adjoining blue photo site’s values and average them, i.e. (b1+b2/2) = B (for target pixel);

For same target pixel, take top and bottom adjoining red photo site’s values and average them,
i.e. (r1+r2/2)= R (for target pixel);

For next horizontal pixel, repeat same process with the next green diamond horizontally to the right;

And so on.

(that’s a bit of simplification as there’s a lot of signal conditioning etc. etc but the basic data flow and structure perhaps suffices)

No that's definitely incorrect. That method you are proposing would cause some serious aliasing artifacts and color fringing. I would assume they are using something a bit more complicated. Likewise there is no reason why a similar algorithm couldn't give you the same exact results on a bayer pattern that isn't rotated 45 degrees.

Not saying there is no advantage to Q67 for downsampling. Just that you should go deeper into the actual math before you draw your conclusions.

IBloom

 

[Eric Lange]

No that's definitely incorrect. That method you are proposing would cause some serious aliasing artifacts and color fringing. I would assume they are using something a bit more complicated. Likewise there is no reason why a similar algorithm couldn't give you the same exact results on a bayer pattern that isn't rotated 45 degrees.

Not saying there is no advantage to Q67 for downsampling. Just that you should go deeper into the actual math before you draw your conclusions.

IBloom

I like and certainly see what you are saying, but I think you would be surprised.

When you look at a test target and look at an array of photo sites, it’s tempting to believe that your black and white resolution bars just jump off the target and go “splat” against the sensor. In reality a lens can only resolve so far ,most of the contrast disappears even for very good glass around 40 line pairs per millimeter, or basically 12 micron. A typical photo site is going to be between 5 and 7.5 micron. And then on top of that you will have your OLPF cutting the effective resolution further. There’s another consideration about true otpics that a single focusable point actually has a point spread function PSF (a bit like a bell curve) as well an associated (so called) circle of confusion + diffraction related artifacts/airy disk. So the actual resolvable black and white bars that hit the sensor have to be way chunkier than you imagine in relation to the size of an individual photo sites (almost depressingly so). This also dosen’t take into account the signal noise within the system and crosstalk between adjacenet and corresponding photosite. So I understand what you see perhaps in your minds eye like wow this would create tons of aliasing artifact, but I think you would be surprised at what is possible even with a crude algorithm; just keep in mind the relative scale between a given photo site and what in practice (using targets, lenses and OLPFs, what is actually resolvable).

The most basic De-bayer is based on neighborhood averaging technique, you can use bi-linear, (a bit crude and chunky), bi-cubic (smoother result) and a whole host of non-linear functions plus some more interesting and imaginative things being used in Graeme’s special sauce that we can only guess at. But a good cross point neighborhood average is good place to start to get a feel of how these things behave. Then you go from there.

The thing that’s interesting is that they (SONY) really are obsessed it seems with this green diamond, as they say that it contributes 100% of the green value for the resultant pixel (effectively) un-diluted. So patterns that would involve the arrangement of four green as the 100% contribution seem unlikely as the four greens have a red photo site smack right in the middle of the four greens. As we have discussed there are many different ways that these values can be sampled but they seem adamant that it’s using a single central green photo site to contribute “100% green” to the final pixel. The thing is if you try more “fancy” interpolation algorithms then the green in effect becomes mushed up (in their eyes), and they are arguing against this idea.

In other words I think physically and theoretically what they are trying to put forward is the idea that they are a half way house between old uncle Bayer and too difficult RGB Panny.

In the next couple of months I need get back into raster based programming as I have to write a lot of new 3d extraction algorithms this might be one I can slip into the mix for fun.

As someone that actually holds patents on some fairly advanced 3d technology, you would be surprised at the number of new and useful things that have never been tried because they seem very counter-intuitive, if not “crazy”. I think this is why the CTO refers to (and I quote) “this ..crazy algorithm”.


If anybody wants to suggest a supposed SONY style Q67 algorithm for me to try I am all ears. The one I am going to try first to get a baseline will be the cross neighborhood average method + pure green. I do believe them when they say they are not De-Bayering. But it will be interesting to eventually find out what’s really going on?

If it’s in the patent literature, then it would not be beyond the wit of man to dig it up there, but my right to experiment might be substantially curtailed.

Cheers,

Eric

 

[Dj Joofa]

Guys you are just beating a semi-dead horse. Demosaicing is not a new topic. Public domiain demosaic algorithms are pretty effective these days. I'm not sure that a competitor to Red will come in the domain of demosaicing solely, rather, IMHO in the areas of overall image quality, ease of use, price point, delivery schedule, and something that is generally perceived to be "state of the art".

Red had a good timing in the industry with an interesting idea regarding increased resolution at a manageable price point. That is why the industry tolerated an immature camera that progressively became better.

Please reread Mike Most's message and don't worry about demosaic too much.

Joofa

 

[Eric Lange]

Guys you are just beating a semi-dead horse. Demosaicing is not a new topic. Public domiain demosaic algorithms are pretty effective these days. I'm not sure that a competitor to Red will come in the domain of demosaicing solely, rather, IMHO in the areas of overall image quality, ease of use, price point, delivery schedule, and something that is generally perceived to be "state of the art".

Red had a good timing in the industry with an interesting idea regarding increased resolution at a manageable price point. That is why the industry tolerated an immature camera that progressively became better.

Please reread Mike Most's message and don't worry about demosaic too much.

Joofa

Sure , I agree it’s a half dead horse, I like very much what you write and the perspective that you bring, for sure.

The thing that is worrying is the way they are marketing the Q67. Their main thrust and plank IS their new fangled 45 degree 100% green diamond which is claimed to be the same as a true Green from a true RGB system, (which it’s not, it’s only 50%). So what's the big deal?

They are basically saying that they want to bury “old uncle” Bayer, because Uncle Bayer is only 33% and new improved Q67 is 67% . So they are going out of their way to influence decision makers with buying power and potential partners that what they will bring to the table is twice as “good” as “old uncle” Bayer. The problem is that because buyers are not that interested in math and various algorithms that they are happy to accept SONY’s slight of hand when presenting these “facts”. The other thing is that media producers that are wary of going digital over film, have now been given a reason to hold off until SONY comes out with Q67 because it will be more film like (in Gamut. Aliasing etc. etc.); which it probably won’t.

If you just roll with what they said, then as a purchaser you would think, hmmm I’m NOT going to invest in current and emerging systems such as RED etc. etc., and I’m going to hold off until SONY can give me the whole shooting match because they have something that’s twice as good as products that are built around creaky old Bayer.

The problem is, that they are using deceptive and misleading figures and technical arguments as their central plank as to why you should hold off and not invest in digital systems until Q67 is ready. It’s basically a pre-emptive blocking move where claimed technical processes and capabilities are at the heart of the matter.

Most of the people reading this thread probably suspect that Q67 is probably NOT going to have major advantages over RED/Epic/redcode etc. purely in terms of image quality, but the other possible advantages might seem more obvious. However a lot of “decision” makers will just go with the flow, and believe what they are told. Everybody needs to know which horses they need to back in advance. Getting answers to these questions sooner rather that later is very important for certain companies and individuals that are developing companion products and technologies through R&D. It’s a little different if your only decision is what is currently available that I can hire next week.

As much as fiddly algorithms and screwy math seem boring they are using these points as their central argument to gain market share and substantially influence decision makers before the instrument can even be publicly tested.

I admit it’s an odd one,

 

[Craig W. Bickerstaff]

I love this idea that for some reason Sony is out to get bayer pattern, especially when they also make a number of bayer pattern cameras.
Sony are invested in many different sensor technologies CMOS, CCD, Bayer, 3 Chip you name it, this is just a guy saying "We think this could be the correct technology to use for this camera"

 

[paulherrin]

At this point, I have to ask the obvious question: Who the hell cares?

haha. hahaha.

 

[Robin Balas]

Just as a funny idea, you could put the SONY algorithm and method to the test (right now) just about.

1.

Using a CCD shoot various target images with the sensor and camera turned at 45 degrees. Given that the SONY pattern happens to be the same as the layout for a rotated Bayer. (A medium format camera with giant pixels and sharp lenses would be good for this to eliminate other factors, but not essential).

2.

You would have to write a sampling algorithm that reads, filters and processes the RAW image in the way they suggest, i.e. pull the green values directly and average the adjoining blue and red values for the corresponding target (rendered pixel) moving in a rectilinear way hopping ("horizontally") from green diamond to green diamond.

3.

You can end up with a 1k by 1k image that is rectified for normal display.

4.

Compare this with a rectilinear conventional square array of demosiac’d conventional Bayer pattern RAW image that has the same number of RGB photo elements as the SONY.

This is a good project for college student or out of work programmer/technician or or professional enthusiast with time on his hands. Probably not more than three days of programming, or if you’re really good it could be done in one day.

Initially I would predict that there would be a fair amount of tweaking to make the jerry rigged/home brew “Q67” produce imagery that is as good as a conventional de-Bayer.
Rotated rectilinear resolution targets would be fun to try out also.

Why wait?

(maybe a good project for Dan...??)

You could call it the “Q67 Homebrew Challenge”...

Just pick up a fuji finepix S2/S3/S5 dSLR, turn off the extended DR function and snap away. The color filter configuration and sensel positions are identical to SONYS illustrations. Or read the now quite old frustrated posts on the boards over the first issues later to be solved in the 3. generation of the sensor and see the tests that showed that there was very little if any real gain at all over comparable resolution "normal" bayer sensors.
But I find it strange that SONY focus so hard on that topic when the best part of the RED cameras is the raw file workflow, not the sensor. Everything else is secondary to that fact (IMHO).

Its quite odd that the movie cam industry seems to re-invent parts of the old stills photo technology, although in a moving image application, and gives the impression of being pioneers on all fronts. Reading this boards and others is sometimes like a deja vu in some aspects where purists don't want to change a thing and the bold ones pushes new ideas forward, skepticism over new ways of doing post with raw files etc. - whole point is that the same results are getting through here and I guess is becoming a new standard. It did pay to be bold with this technology in the "old" days with stills photo and it looks like it pays off in the moving image business as well. The fact that some of the core idees and some of the technology isn't all that new, as it have succeeded once before, should help people see that.

Now give me that 645 sensor cam - I have been waiting for a long time And I guess I will be waiting some more... :Yawn:

 

[Eric Lange]

Just pick up a fuji finepix S2/S3/S5 dSLR, turn off the extended DR function and snap away. The color filter configuration and sensel positions are identical to SONYS illustrations. Or read the now quite old frustrated posts on the boards over the first issues later to be solved in the 3. generation of the sensor and see the tests that showed that there was very little if any real gain at all over comparable resolution "normal" bayer sensors.
But I find it strange that SONY focus so hard on that topic when the best part of the RED cameras is the raw file workflow, not the sensor. Everything else is secondary to that fact (IMHO).

Its quite odd that the movie cam industry seems to re-invent parts of the old stills photo technology, although in a moving image application, and gives the impression of being pioneers on all fronts. Reading this boards and others is sometimes like a deja vu in some aspects where purists don't want to change a thing and the bold ones pushes new ideas forward, skepticism over new ways of doing post with raw files etc. - whole point is that the same results are getting through here and I guess is becoming a new standard. It did pay to be bold with this technology in the "old" days with stills photo and it looks like it pays off in the moving image business as well. The fact that some of the core idees and some of the technology isn't all that new, as it have succeeded once before, should help people see that.

Now give me that 645 sensor cam - I have been waiting for a long time And I guess I will be waiting some more... :Yawn:

Yup, the Octagonal array is kinda cute, that's interesting what you said about the third generation of sensor. I’m definitely Jonesing for the 645 too, but that may be way into the future given everything else that’s going on.

 

[Eric Lange]

I love this idea that for some reason Sony is out to get bayer pattern, especially when they also make a number of bayer pattern cameras.
Sony are invested in many different sensor technologies CMOS, CCD, Bayer, 3 Chip you name it, this is just a guy saying "We think this could be the correct technology to use for this camera"

Talk to Graeme about that, it’s not just me being super overly paranoid…

I agree, the world of 4k delivery is a bit myopic compared to the universe of other “gadgets”. I would certainly agree with what you are saying in terms of tone in the presentation as being utterly reasonable up until about 7 mins 30s into the presentation where everything (for me at least) suddenly takes a sudden lurch to the left. That’s where everything starts to go pear shaped and then after five minutes everything turns back to normal when he starts talking about groovy tape decks.

 

[Sanjin Jukic]

Sure , I agree it’s a half dead horse, I like very much what you write and the perspective that you bring, for sure.

The thing that is worrying is the way they are marketing the Q67. Their main thrust and plank IS their new fangled 45 degree 100% green diamond which is claimed to be the same as a true Green from a true RGB system, (which it’s not, it’s only 50%). So what's the big deal?

They are basically saying that they want to bury “old uncle” Bayer, because Uncle Bayer is only 33% and new improved Q67 is 67% . So they are going out of their way to influence decision makers with buying power and potential partners that what they will bring to the table is twice as “good” as “old uncle” Bayer. The problem is that because buyers are not that interested in math and various algorithms that they are happy to accept SONY’s slight of hand when presenting these “facts”. The other thing is that media producers that are wary of going digital over film, have now been given a reason to hold off until SONY comes out with Q67 because it will be more film like (in Gamut. Aliasing etc. etc.); which it probably won’t.

If you just roll with what they said, then as a purchaser you would think, hmmm I’m NOT going to invest in current and emerging systems such as RED etc. etc., and I’m going to hold off until SONY can give me the whole shooting match because they have something that’s twice as good as products that are built around creaky old Bayer.

The problem is, that they are using deceptive and misleading figures and technical arguments as their central plank as to why you should hold off and not invest in digital systems until Q67 is ready. It’s basically a pre-emptive blocking move where claimed technical processes and capabilities are at the heart of the matter.

Most of the people reading this thread probably suspect that Q67 is probably NOT going to have major advantages over RED/Epic/redcode etc. purely in terms of image quality, but the other possible advantages might seem more obvious. However a lot of “decision” makers will just go with the flow, and believe what they are told. Everybody needs to know which horses they need to back in advance. Getting answers to these questions sooner rather that later is very important for certain companies and individuals that are developing companion products and technologies through R&D. It’s a little different if your only decision is what is currently available that I can hire next week.

As much as fiddly algorithms and screwy math seem boring they are using these points as their central argument to gain market share and substantially influence decision makers before the instrument can even be publicly tested.

I admit it’s an odd one,

Eric, well said.

 

[Eric Lange]

No that's definitely incorrect. That method you are proposing would cause some serious aliasing artifacts and color fringing. I would assume they are using something a bit more complicated. Likewise there is no reason why a similar algorithm couldn't give you the same exact results on a bayer pattern that isn't rotated 45 degrees.

Not saying there is no advantage to Q67 for downsampling. Just that you should go deeper into the actual math before you draw your conclusions.

IBloom

…And yes (as you say) you can carry that out on a non-rotated Bayer pattern, there are several patents (at least) that I have stumbled across in a short period of time that deal with exactly that!

IBLOOM, this might be more your cup of tea

US Patent 5,629,743 (may 13, 1997)

www.google.com/patents?hl=en&lr=&vi...lements&printsec=abstract#v=onepage&q&f=false

(Its pretty well explained)

 

[Eric Lange]

So this patent application describes why having photo sites oriented at 45 degree angles makes for a pretty fast zig-zag read out.

Cambel Dec 2nd, 2003 US patent 6,657,755

www.freepatentsonline.com/6657755.pdf

I would really encourage people to try and read the patent literature as generally things are VERY accessible and quite often well explained. Most of the technical subjects have to be substantially “dumbed” down so that the patent examiner can easily understand what is really going on; its very difficult as an examiner when you have to encompass several technical fields to a deep level. Also Patents have to be clear and understandable in terms of simple claim language that will have to stand up in a court of law in an infringement suit judged by a jury pulled from the general public.

check it out...