Reading mode: Light Dark

Sensor ?

Go to last post
Roland Karlsson said:
Hmmmm ... this was a faulty example IMHO. The strange look is due to
excessive noise reduction. No low noise CFA images looks anything
like this.
Click to expand...
It's from an 8MP 1/2.3" sensor at ISO 800. Imagine what an X3 equivalent 4x3 MP 1/2.3"sensor (e.g. a pixel pitch of 2.5um) would look like at ISO 800.

--
Erik
 
1chaz said:
If the sd 14 is considerd a 4.7 mp sensor (capturing 3 colors at each
pixel ) then wouldn"t a 12 mp beyer sensor really be a 4.0 mp X 3
colors = 12 ???.. I have a new sd 14 with a 18x50 ex and also own a
10d and a 20d and have taken identical shots and can see why people
love this sensor yes the body is not as perfected or polished as a
canon or nikon but that iq is impressive and the color is spot on.
Such large files for a 4.7 mp lots o bits
--
Click to expand...
--

I really don't like how the term pixels are used in sensors. I would be much happier if they would use the same term as for monitor. One pixel is an array of 3 sub pixels (red, green, blue). See wikipedia: http://en.wikipedia.org/wiki/Pixel

Then a Sigma sd14 would be a 4.7 Mpixel camera with 3 subpixels per pixel. A 14Mpixel camera would be 3.5 Mpixel camera with 4 subpixels per pixel (2 green, 1 red, 1 blue).

I think it would also be more honest towards the customer as monitors are also sold with one pixel holds all primary colours i.e. 3 sub pixels. No one would except a LCD monitor that would be sold with the count of the sub pixels instead of the pixel.
 
Joop Boonen said:
I really don't like how the term pixels are used in sensors. I would
be much happier if they would use the same term as for monitor. One
pixel is an array of 3 sub pixels (red, green, blue). See wikipedia:
http://en.wikipedia.org/wiki/Pixel

Then a Sigma sd14 would be a 4.7 Mpixel camera with 3 subpixels per
pixel. A 14Mpixel camera would be 3.5 Mpixel camera with 4 subpixels
per pixel (2 green, 1 red, 1 blue).
Click to expand...
That is what I would like too. So say G9 would be advertised as 12 mega sub-pixels.
Dp1 would be 14 mega sub-pixels.

Counting 4 subpixels as one pixel in Bayer CFA is a bit problematic though since they contribute to neighbours too.
Joop Boonen said:
I think it would also be more honest towards the customer as monitors
are also sold with one pixel holds all primary colours i.e. 3 sub
pixels. No one would except a LCD monitor that would be sold with the
count of the sub pixels instead of the pixel.
Click to expand...
I would not be happy if my 1600x1200 LCD monitor would have every pixel of only single (r,g,b)-color.
 
Joop Boonen said:
I really don't like how the term pixels are used in sensors. I would
be much happier if they would use the same term as for monitor. One
pixel is an array of 3 sub pixels (red, green, blue). See wikipedia:
http://en.wikipedia.org/wiki/Pixel
Click to expand...
A monitor is a bad example. Both methods of counting pixels are used. And there are several configurations how color is created - at least for the CRT monitors. For CRT monitors you have DPI and not PPI - because the color dots are not aligned to the pixels at all.

Use the example of a color image file instead - then you get the right notion.
Joop Boonen said:
Then a Sigma sd14 would be a 4.7 Mpixel camera with 3 subpixels per
pixel. A 14Mpixel camera would be 3.5 Mpixel camera with 4 subpixels
per pixel (2 green, 1 red, 1 blue).
Click to expand...
I agree fully with the Sigma. But dividing Bayer with 3 is strange IMHO. There exist no natural divisor. So - in practice 1 have to be used.

This makes it impossible to compare Foveon and Bayer using pixel counting, which is a good thing IMHO.
Joop Boonen said:
I think it would also be more honest towards the customer as monitors
are also sold with one pixel holds all primary colours i.e. 3 sub
pixels. No one would except a LCD monitor that would be sold with the
count of the sub pixels instead of the pixel.
Click to expand...
Yet they do. Care to try to count the pixel of your LCD screen on your camera?

--
Roland
 
Erik Magnuson said:
It's from an 8MP 1/2.3" sensor at ISO 800.
Click to expand...
No it is not - at least not unmodified. It is heavily smoothed - like using some artistic filter in Photoshop. I assume that the original image is very noisy - but sharp. I would rather have this to start with and do my own noise reduction.
Member said:
Imagine what an X3
equivalent 4x3 MP 1/2.3"sensor (e.g. a pixel pitch of 2.5um) would
look like at ISO 800.
Click to expand...
What was the pitch for the "polaroid" camera? Was it 5 um?

--
Roland
 
Roland Karlsson said:
Then you have to take the conversion algorithm into account.
According to a white paper on Foveon´s web pages - the conversion is
based on using the middle layer for luminosity and then extracting
downsampled difference channels for color information.
Click to expand...
As far as I remember it this white paper clearly states that this approach was indented for fast processing with not so much computation power while high end cameras and PC RAW conversion software would use different methods.

Imho this is clearly aimed at compact DSC / cell phone applications and not SLRs.

Furthermore consider that this article is much newer than the SD9 or SD10, and as we know the SD14 raw files still worked with the old raw processors (just with less heavy noise reduction), so it seems very unlikely that they actually used this in any of the current cameras.

And btw. another key aspect is that with the method proposed in the paper one starts with "full" sampling of luma and chroma information and then blurs the chroma a bit. The impact of this is quite different from starting with a incomplete sampling of luma and chroma and interpolating most of it.

--
http://www.pbase.com/dgross

 
Dominic Groß said:
As far as I remember it this white paper clearly states that this
approach was indented for fast processing with not so much
computation power while high end cameras and PC RAW conversion
software would use different methods.
Click to expand...
Actually - I dont think so. The paper is slightly fuzzy about this. Yes - it says that the method is suitable to use in smaller cameras. But - it also states that the downsampling is important for noise reduction. So ...
Dominic Groß said:
Furthermore consider that this article is much newer than the SD9 or
SD10, and as we know the SD14 raw files still worked with the old raw
processors (just with less heavy noise reduction), so it seems very
unlikely that they actually used this in any of the current cameras.
Click to expand...
I dont understand your reasoning there. How can the fact that the paper is new make it likely that new cameras do not use whats in it? And ... the format of the RAW files has nothing to do with the conversion from RAW to RGB.
Dominic Groß said:
And btw. another key aspect is that with the method proposed in the
paper one starts with "full" sampling of luma and chroma information
and then blurs the chroma a bit. The impact of this is quite
different from starting with a incomplete sampling of luma and chroma
and interpolating most of it.
Click to expand...
Yes - thats right. I just wanted to give the full picture. Some thinks Foveon images are non AA full native RGB images - without any trace of blurring except what the lens does. They are not.

--
Roland
 
Roland Karlsson said:
Now - I think you are trying to be difficult - and choose not to
understand what I mean :)
Click to expand...
No, I'm emphasizing that anything coming out of X3 processing (particularly in-camera processing) for similar sensor size will also be heavily modified.
Roland Karlsson said:
But - what we saw there was heavily modified.
Click to expand...
Of course. No one has yet to get good per-pixel detail out of tiny pixels.

--
Erik
 
Erik Magnuson said:
Of course. No one has yet to get good per-pixel detail out of tiny
pixels.
Click to expand...
In good lighting my Fuijicolor f20 makes rather nice images - but when it gets dimmer - then the image looks like an acrylic painting.

--
Roland
 
Roland Karlsson said:
Dominic Groß said:
As far as I remember it this white paper clearly states that this
approach was indented for fast processing with not so much
computation power while high end cameras and PC RAW conversion
software would use different methods.
Click to expand...
Actually - I dont think so. The paper is slightly fuzzy about this.
Yes - it says that the method is suitable to use in smaller cameras.
Click to expand...
As far as I remeber it it describes that raw processing so far amplifying noise because of the RGB conversion and therefore requires advanced noise reduction.

It also says that for RAW processing such multipass methods are used. So I don't really see where to read into this that this is used for the SD14 / DP1.
Roland Karlsson said:
Roland Karlsson said:
Furthermore consider that this article is much newer than the SD9 or
SD10, and as we know the SD14 raw files still worked with the old raw
processors (just with less heavy noise reduction), so it seems very
unlikely that they actually used this in any of the current cameras.
Click to expand...
I dont understand your reasoning there. How can the fact that the
paper is new make it likely that new cameras do not use whats in it?
Click to expand...
There is also some evidence that they don't use it. First of all processing takes even longer for the SD14 than previous cameras, suggesting that there is even more processing. The resulting images for the SD14 look like the old processing with better noise reduction on top. etc.
Roland Karlsson said:
Yes - thats right. I just wanted to give the full picture. Some
thinks Foveon images are non AA full native RGB images - without any
trace of blurring except what the lens does. They are not.
Click to expand...
It is interesting that you state this as fact when you have no hard evidence.
To me all this seems very unlikely for the SD14 / DP1.

Should be possible to test this with a synthetic raw file...

--
http://www.pbase.com/dgross

 
Dominic Groß said:
Roland said:
Yes - thats right. I just wanted to give the full picture. Some
thinks Foveon images are non AA full native RGB images - without any
trace of blurring except what the lens does. They are not.
Click to expand...
It is interesting that you state this as fact when you have no hard
evidence.
Click to expand...
Hmmmm ... is it a native RGB sensor?

OK - OK - thats nitpicking :) You meant that I have no hard evidence that they do chroma blur. But - I meant exactly what I wrote (including the RGB part) and I do not want to state 100% that they do any chroma blurring - because I do not know that.

I would be VERY surprised if the noise reduction they DO have do NOT include any kind of chroma blurring.
Dominic Groß said:
To me all this seems very unlikely for the SD14 / DP1.
Click to expand...
Hmmmm ... unlikely that they use (1) the two pipeline method (2) chroma downsampling and/or (3) chroma blurring?
Dominic Groß said:
Should be possible to test this with a synthetic raw file...
Click to expand...
And - as a matter of fact - I do have a program that easily can be adapted to generate such a synthetic file :) What test pattern do you suggest?

Just a caveat there. The RAW compression method Sigma uses is not non lossy. It is based on progressive delta coding - and some ringing might be possible at edges. (But to come to think of it - if all lines are horizontal - then the coding will be non lossy.)

--
Roland
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top