Go Back   Steve's Digicams Forums > Digicam Help > General Discussion

Reply
 
Thread Tools Search this Thread
Old Apr 1, 2006, 4:57 PM   #1
Junior Member
 
Join Date: Apr 2006
Posts: 8
Default

There are a lot of discussions about the noise that results from cramming too many Pixels in a small chip. In threads about the FZ-30 I read the canera would have been better if Panasonic had only put 6 MPixels on the chip. That got me wondering. It is obvious that a camera with 8 MPixel on the same chip size as a 5 MPixel camera has a worse signal to noise ratio because each pixel gets only 5/8th of the light. That would be obvious when watching the pictures at 100%. However that seems to be an unfair comparison since the 8Mpixel camera also has more Pixel. When viewing full-screen on a computer screen with resolution 1600x1200 the pictures of both cameras get down-converted. How does this affect the noise levels? Down conversion cancels out some noise, but to what extent? My assumption would be, that the 8 MPixel camera has still a bit more noise than the 5 MPixel camera, but that the effect is much much less obvious than when watching at 100 percent. I somehow can't figure out how to average signal to noise rations and I am too lazy too look up the statistics.

Are there any cameras on the market where a comparable sensor size and technique is used with different resolutions?
uz-guy is offline   Reply With Quote
Sponsored Links
Old Apr 2, 2006, 12:23 PM   #2
Senior Member
 
Join Date: Dec 2002
Posts: 5,803
Default

I know of no compairson like you are looking for.

You are correct, though, reducing the image (or viewing it at less than 100%) will both mask detail and noise. So you will see less of it. The problem, in my experience, is that you will probably want to add some sharpening to the image when you finish (at least with a DSLR you do, with point-and-shoot cameras they come out-of-the-camera with more sharpening.) And when you do, the noise will look even worse.

I find that I have to be very careful when sharpening images to make sure the noise doesn't become distracting.

Eric
eric s is offline   Reply With Quote
Old Apr 2, 2006, 2:44 PM   #3
Junior Member
 
Join Date: Apr 2006
Posts: 8
Default

eric sYou are correct, though, reducing the image (or viewing it at less than 100%) will both mask detail and noise.* So you will see less of it.*
[/quote]
See, I am not interested in how to best denoise a picture, I know there are lots of posts on this issue and several programs that supposedly do a good job. I am more interested in the theoritcal background: Does a camera with more Pixels on exactly the same chip produce worse pictures. Can more Mpixels ever be a disadavtage or are they just no longer a benefit cause the noise kills the advantage of the higher resolution.
I wonder whether there is a way to estimate this. There would have to be a way to estimate the statistical nature of camera noise and how it behaves when averaging several signals while downsizing.
uz-guy is offline   Reply With Quote
Old Apr 2, 2006, 2:59 PM   #4
E.T
Senior Member
 
E.T's Avatar
 
Join Date: Jul 2003
Posts: 921
Default

uz-guy wrote:
Quote:
I am more interested in the theoritcal background: Does a camera with more Pixels on exactly the same chip produce worse pictures...
I wonder whether there is a way to estimate this.
Estimate that from fact that current high megapixel cameras need to use lot of noise reduction while say three years ago noise reduction was used much less.
E.T is offline   Reply With Quote
Old Apr 3, 2006, 4:06 PM   #5
Junior Member
 
Join Date: Apr 2006
Posts: 8
Default

E.T
Quote:
Estimate that from fact that current high megapixel cameras need to use lot of noise reduction while say three years ago noise reduction was used much less.
That doesn't say much because people have much higher demands on a 8 MPixel camera than on a 3 Mpixel camera and look at the pictures at 100 %. I have no doubt that the noise per pixel is higher with the 8 MPixel camera, but I am wondering what happens when you resize both pictures to the same size, for example when doing a print.
uz-guy is offline   Reply With Quote
Old Apr 3, 2006, 10:55 PM   #6
Senior Member
 
mtngal's Avatar
 
Join Date: Nov 2005
Location: Frazier Park, CA
Posts: 16,052
Default

I can't give you any mathmatical formula for such things,only mypersonal experience (basically a beginner, so this was all unscientific). I took a number of pictures with a 5 MP Sony F717 (faithful camera for 3 years but it developed some intermittent problems) and a new8 MP Panasonic FZ30 that I bought before Christmasto replace it. I took a number of photos and reduced them (not cropped)to the same size (the 8 MP files were reduced by a greater percentage than the 5 MP - they were both reduced to the size of my monitor). The 3 year old Sony consistentlytook better pictures, it had more dynamic range so it captured more detail in the shadows. After that experience, I concluded that more pixels do not mean better quality, sold the Panasonic after Christmas for a big loss and bought a 6 MP Pentax *istDS (should have done that in the first place).
mtngal is offline   Reply With Quote
Old Apr 4, 2006, 9:19 AM   #7
Senior Member
 
Join Date: Dec 2002
Posts: 5,803
Default

uz-guy,

Ah, ok. I like thinking about things and looking for trends/conclusions.

Can it be "bad" to have more MP? Humm....
As far as noise goes, it could be worse if everything else is equal (the problem is that it never is.) The digital sensor is made up of things called a "photosite", which is what actually measures the light and turns it into electrical impulses. It really isn't that the MP is higher that it the problem, the problem is that the photosites are smaller (to let you put more on the same size sensor.)

If you have two sensors that are the same size, then as you increase the MP the photosites must either become smaller or be packed closer together. The problem is that the closer you pack photosites together the more they can interfere with each other (electrically) and this interference is one of the source of noise in the image.

Another problem with the smaller photosites is that they capture less light (fundamentally, less light hits a smaller area than a larger one.) Therefor other changes would have to be made to make the photosite more sensitive or the sigal from them has to be boosted. Both of those things can also increase the electrical "noise" in the system, which increases the noise in the image.

Of course, no camera maker "only" increases the number of photosites on the sensor. They also get better at suspressing electricial interference (which lets them boost photosite sensitivity) and other things to reduce noise.

It is a constant give-and-take, though. As you add more MP (without increasing surface area) you have to do other things to reduce the noise you just increased. That is why there is really no "formula" to map out the problem. There are so many variables, so many things that change between cameras beyond just photosite size, that it really can't be represented that way.

As an aside....
Another problem is that if the sensor is physically big enough and the photosites small enough, then micro mirrors have to be added to the sensor. The problem is that photosites are more sensitive to the angle the light hits them than film is. The angle the light hits in the center of the sensor is near purpendicular. But as you move further to the edge it hits at a steeper and steeper angle. This is bad, and requires money (the cost of the micro mirrors) to fix. Smaller photosites means the light has to be closer to purpendicular to be useful... so that means more micro mirrors and a more costly camera. They don't effect noise, if done correctly, but it is an effect of smaller sensors. This is one reason why the sensor on the Canon 1D was so expensive, as far as I know it was the first camera to use a sensor that size and the need for micro mirrors was a new problem.

Eric
eric s is offline   Reply With Quote
Old Apr 4, 2006, 11:14 AM   #8
Senior Member
 
Join Date: Sep 2005
Posts: 1,093
Default

uz-guy wrote:
Quote:
...When viewing full-screen on a computer screen with resolution 1600x1200 the pictures of both cameras get down-converted. How does this affect the noise levels? Down conversion cancels out some noise, but to what extent? My assumption would be, that the 8 MPixel camera has still a bit more noise than the 5 MPixel camera, but that the effect is much much less obvious than when watching at 100 percent. I somehow can't figure out how to average signal to noise rations and I am too lazy too look up the statistics.

Are there any cameras on the market where a comparable sensor size and technique is used with different resolutions?
Eric gave an excellent answer that is much more precise than what I am about to say. However, I think that what I will say is more directly an (inexact) answer to what you asked. If you want to know the truth, read Eric. If you want to know the usual wrong answer, keep reading me!

The usual rule of thumb for how white noise S/N improves with samples is that it improves with the square root of the number of samples. If your display software used the excess pixels intelligently, it would reduce the noise by a factor of two if it downsized the image by a factor of four. Thus, the relatie noiseimprovement due to more pixels in an 8MP image vs a 5 MP image after downsizing would be sqrt(8/5), or about 25%. To all intents and purposes, the images will look the same wrt noise -- you want whole number multiples to realy be able to see a difference on these things.

As to your second question, the FZ20/FZ30 closely approximates the test you may have had inmind: as I understand it, the FZ30increased the number of pixels from 5 MP to 8 MP while increasing the sensor size proportionately. The processing applied would be essentially the same for both cameras, as they use the same processor. So you might assume that the noise profile would be the same per pixel in each camera. However, the lens had to be changed to accommodate the larger chip size, etc. So you're back in Eric's world of complicated real-world factors even here.


tclune is offline   Reply With Quote
 
Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



All times are GMT -5. The time now is 6:24 PM.