Doug Morgan wrote:
Aren't you showing the Foveon at 3x the resolution of the bayer?
Not by what Bayer advocates say. Both of my virtual sensors have the same x,y axis. But to be fair the virtual Foveon sensor has 3x the pixels, yes. So does the layered pixels count? You tell me.
mt-m wrote:
All this will only make any sense if you describe the process step-by-step.
Would you please describe how you "made a very low resolution virtual Foveon and Bayer sensor as the sensors would see firsthand"?
The Foveon was easy since it captures 1 to 1 to begin with. The Bayer sensor was harder. I had to take the image and break it down into it's RGB components as how the Bayer sensor is laid out on it's single layer, took awhile to do. I just used it's 25% RED 25% BLUE AND 50% GREEN photosite array as illustrated all over the web.
However, as Doug points out, to complete the analogy you should now do 2x and 3x bayer pixels to see where the images ar similar in detail (even though it may be that at that poibnt the Foveon gets the absolute colors a t a pixel more correct.!)
I am sorry that foveon technology is not open since it would take the bayer vs foveon debate out of the picture - let the camera makers decide on the best approach for their markets.
Also, you should recognize that the bayer matrix interpolations are highly proprietary, and can involve complex algorithms for high and low bandwidth luminance an chrominance sections, for edges, for black etc. It is why RAW converters vary so greatly.
But I am impressed with the straightforwardness of the model. Great work.
Now, can you do the 2x ad 3x? My bet is that(except for the truer color a t a pixel, foveonedge, almost de facto) that the details will converge somewhere in between these values.
Larry,
It seems that you are trying to make your point that Foveon is inherently superior to Bayer. We all got your point ! Some may not agree with you. But everyone understands your position. Repeating is not necessary unless there is new data.
If I were you, I would go out and enjoy talking photos with the superior sensor.
Regardless the appearanceof 'lobbying' in these discussion, I find Larry's contribution to an 'evidenced based' discussion very refreshing. Fan is one thing, facts are another. This is exactly the kind of price-performance modeling I would expect camera makers to go through in their evaluation of what sensors to use. Ther may in fact be hidden issues that will constrain larger MF foveon. But we should all be open minded when a solid piece of work has been done, even by a fan of one or other appraoches.
Guys guys... i'm not trying to rain on your parades... There are lots of people that don't realize how much technology that Bayer has to go through to get great images and lots of people are confused on pixel count of Foveon and just wanted to illustrate that those hidden pixels of Foveon are there. Please settle down I do hope that Bayer is gone in the next 10 years in high end imagery but that is only my opinion. I think we all have lots of things to look forward to.
Larry, I am with the others. I don't think there is anyone here who doesn't agree that the Foveon sensor is better than a Bayer sensor, pixel for pixel, but this isn't reality. The SD14 gives 4.7 million (great) pixels, and a similarly priced camera (guessing here) gives perhaps 10 million (poor) pixels. The logical thing would be to compare what is on the market, ie. give the Bayer images 2.13x times the resolution, and see where that leaves the results.
The two images have the same amount of resulting IMAGE-pixels: 302 x 446 pixels.
So the two images have the same concrete resolution.
The Foveon has an advantage in accurate color rendition because it used 3 photodetectors (or SENSOR-pixels) in three layers to capture light and produce every single outgoing, resulting 1 IMAGE-pixel.
That is why the Foveon based Sigma SD14 camera uses as much as 14 million photodetectors (or SENSOR-pixels) in three layers to produce only 4,7 million outgoing, resulting IMAGE-pixels.
I can easily understand that the use of 3 SENSOR-pixels (red, green, blue) to produce 1 resulting IMAGE-pixel gives the Foveon sensor an advantage in qualitywise rendering of colors in these "few" resulting IMAGE-pixels. They really ought to have a very high standard of accuracy regarding the "consumption of photodetectors".
I can also understand that the accurate rendering of a slight difference between two resulting neighbour IMAGE-pixels might very well provide you with an impression of better micro contrast and sharpness and perceived resolution, even though we compare two sample images of exactly the same amount of resulting IMAGE-pixels: 302 x 446, i.e. same concrete resolution.
And I can also understand that a clean image of such accurate color IMAGE-pixels might do very well in upsizing by interpolation.
Furthermore I like the idea of avoiding the Bayer way:
in-camera interpolating >> makes color artifacts >> to be reduced by blur filters >> with loss of detail as a result.
What I don't like too much is the thought of maybe often having to upsize images by interpolation in postprocessing. Today 4,7 million resulting IMAGE-pixels really isn't that much.
So, Larry, do you know the reason why Foveon / Sigma don't go for a "SD30" camera producing resulting images of around 10 million IMAGE-pixels just like all their nearest competitors, (only of course with Foveons better quality of rendering each single image-pixel) ?
Do you think it is at this stage:
(a) technically too difficult to manufacture getting an acceptable percentage through the final quality control ?
(b) too expensive to produce ?
(c) or would it imply too many informations to process, is the processor the bottleneck ?
Anyway, Larry, you have really made me curious about this Foveon sensor technology, it seems so obvious and I need to try the SD14 some day. I hope we will see a lot of your images as soon as you get your SD14.
And that goes for you too, Grant.
I would also like to see this sensor type in some other camera brands. Who knows, it might even become the rescue for the Four-Thirds Standard with their scaring small sensor size ?
Before you wrote about the Foveon in the other threads I knew less than nothing about it. I think it is very interesting and exciting with such new innovative solutions on sensor technology. That is really what we need. Thanks for all your work with sharing.
Kind regards - Steen
Edited by Steen Bondo on Oct 01, 2006 at 10:00 AM GMT
carstenw wrote:
Larry, I am with the others. I don't think there is anyone here who doesn't agree that the Foveon sensor is better than a Bayer sensor, pixel for pixel, but this isn't reality. The SD14 gives 4.7 million (great) pixels, and a similarly priced camera (guessing here) gives perhaps 10 million (poor) pixels. The logical thing would be to compare what is on the market, ie. give the Bayer images 2.13x times the resolution, and see where that leaves the results.
Hi carstenw and Steen,
Well, I was one of the Bayer users for almost 4 years? I think close to that maybe more. I was fortunate to have used the best in the 1D line. I was very happy but things started to change when I did more and more images. We start to pay attention to the images more as we do more, I think we can all agree with that. Well the more I looked at the images the more questions I had and the happy feeling started to wain. Then I decided to venture out to see what others had to offer and ran across the Foveon, which was tucked away due to poor advertising and all the bad talk from people who had no earthly idea what they were talking about even though they never even owned the Foveon camera, so it never caught on. Then someone sold me the SD10 and I was skeptical then, but after using the camera it changed my view on what a digital image should be. So now I feel I have to at least share my findings to make the Sigma known to be best of my abilities to at least let people know there are alternatives. I'm sure there are people just like me who aren't so happy and would like to know about other ways instead of being lost in the sea of Bayer.
On the SD14.... I wish I could tell you what I know one word... awesome.
Larry, I totally agree that the Sigma SD14 is a fascinating camera, and I will be watching what happens with it very carefully, and considering what it would mean for me to buy one, but I still do not find it fair to compare pixel-to-pixel, because that isn't the decision you make on the market today, or when the SD14 comes out. I would love to play with one though, and print out some large prints to compare, say A2.
Btw, I would think that the Fuji, Kodak (not sold any more, I know) and Leica DMR/M8 also provide much better images than, say, a Canon, at the pixel level. I would be very curious to see real-world comparisons between the SD14, M8 and S5 Pro, at the pixel level, and overall.
carstenw wrote:
...
, I would think that the Fuji, Kodak (not sold any more, I know) and Leica DMR/M8 also provide much better images than, say, a Canon, at the pixel level. I would be very curious to see real-world comparisons between the SD14, M8 and S5 Pro, at the pixel level, and overall.
My observation is that the "pixel-level" quality has much to do with AA-filter strength.
My D30 ( the very original one ) has better "pixel-level" quality than my 1DsII. The "Pixel-level" look is determined by the pixel pitch relative to the net resolution of the lens and the AA filter. In other words, 1 MP FF camera would give you almost Lego-block level clarity on each pixel.
I have not seen a single photo of SD14. BUT I can predict one thing. Its pixel-level quality will be poorer than SD10.
But I won't!
I do hope that Bayer is gone in the next 10 years in high end imagery but that is only my opinion. I think we all have lots of things to look forward to.