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

Reply
 
Thread Tools Search this Thread
Old Apr 5, 2009, 12:29 PM   #21
Super Moderator
 
peripatetic's Avatar
 
Join Date: Nov 2004
Posts: 3,599
Default

Greg,

I believe that you are substantially correct in your evaluation of this camera.

You do put forward some good points:

1. More pixels => more data, slower throughput, etc. This may work to the detriment of what an individual may require and make for a less usable camera.

2. The fact that with these very high densities the lenses are diffraction limited even at maximum aperture means dramatically diminishing returns on adding extra pixels.

3. And of course on cheap cameras the quality of the lenses and construction means that the sensor is not the weakest link in the imaging chain anymore, therefore adding extra pixels once again has little positive impact on image quality.

So for the small-sensor P&S market the major driving factor behind the increase in pixels is more likely to be marketing than an increase in image quality.

However there is a prevalent myth that DR and Noise are increased and overall image quality reduced by increasing the pixel count, and that is not the case.

Dynamic range, like noise, is unaffected by increasing the pixel count if you stop (and people really should) believing that the arbiter of quality is the 100% view in photoshop.

At the standard output size the downsampling compensates for the decrease in DR at the per pixel level.

So once again you do not lose anything by adding pixels, and you gain resolution.

For a given level of sensor technology, DR is primarily determined by how much light you capture, and therefore (and once again) the only way to make significant gains is to increase the size of the sensor.

On the question of the prevalence of these myths however...

You misunderstand the difference between theory and practice. A theory is something that explains the evidence. If the evidence is contrary, then there is something wrong with the theory.

There are two theories here in competition: one says that adding extra pixels has an adverse affect on image quality (this is the theory you are subscribing to), the other theory says that it does not.

The question is what does the evidence show? In order to trust evidence you rely to some degree on experts who make their own examinations and write that up in the form of articles. The best sources are those like DXO who tell you what their methods are and invite you to run the experiments for yourself. Unfortunately gathering the evidence in a non-biased way is expensive and requires rigorous application of experimental methodology. Anecdotal evidence is not reliable; this is true of all science.

So on the one hand we have people like Ctein who is a reknowned physicist and acknowledge as one of the finest printers in the world. We have evidence from DXO labs, one of the foremost labs in the world, and who publish their methodologies and inviite others to repeat or refute their results. We have additional evidence from other internet sources which one must accord greater or lesser weighting depending on how reliable and convincing you find them.

And other the other hand we have statements like:

Quote:
I will not be convinced differently, even with another dozen articles.
As for me - if you show me better evidence I would be delighted to change my views, and will have learned something in the bargain.

The reader of course, as always, must make up his own mind.





peripatetic is offline   Reply With Quote
Old Apr 5, 2009, 4:44 PM   #22
Senior Member
 
Join Date: Feb 2007
Location: Wisconsin
Posts: 1,241
Default



Thats quite a post and I need to break it into sections to answer.
Quote:
Greg,

I believe that you are substantially correct in your evaluation of this camera.

You do put forward some good points:

1. More pixels => more data, slower throughput, etc. This may work to the detriment of what an individual may require and make for a less usable camera.

2. The fact that with these very high densities the lenses are diffraction limited even at maximum aperture means dramatically diminishing returns on adding extra pixels.

3. And of course on cheap cameras the quality of the lenses and construction means that the sensor is not the weakest link in the imaging chain anymore, therefore adding extra pixels once again has little positive impact on image quality.
That in total is all I said. At least its all I intended to say. That camera would be a better, more useable camera with less pixels. That is certainly my opinion. You will not convince me differently. Funny part is I think in your heart of hearts, you agree with me.

Quote:
However there is a prevalent myth that DR and Noise are increased and overall image quality reduced by increasing the pixel count, and that is not the case.

Dynamic range, like noise, is unaffected by increasing the pixel count if you stop (and people really should) believing that the arbiter of quality is the 100% view in photoshop.

At the standard output size the downsampling compensates for the decrease in DR at the per pixel level.

So once again you do not lose anything by adding pixels, and you gain resolution.
I'll agree with TCav here. Its difficult to compare apples to apples.

I generally agree with the noise arguments. The only reason there is more noise is because there is more data as resolution increases. I didn't address that in my prior posts. I said there was too many pixels in that particular camera.

As to Dynamic range, I don't agree with you at all. I must say I don't know what "dynamic range on a per pixel level" actually means. Fill capacity/256?

I'll admit your post sent me scrambling. I was wondering if I misunderstood the relationship between well capacity and dynamic range. I went to a primary source of yours to find the answer, with as close of an apples to apples comparison as I think exists. DXO comparisons with D3 and D3x with the D90 thrown in to boot. Click on the "Dynamic Range" button.

http://www.dxomark.com/index.php/eng...(brand3)/Nikon

I don't think my understanding of DR is faulty at all. The concession of lowering the gain of the D3x to 100-1600 vs the D3's 200-6400 would tend to support the lower resolution (and older) D3 sensor has higher DR at a given ISO. The curve of the D3 is clearly higher than the D3x.

After doubting my understanding, I've decided to stick with my statement. DR decreases as pixel count increases.
Quote:
You misunderstand the difference between theory and practice. A theory is something that explains the evidence. If the evidence is contrary, then there is something wrong with the theory.
I have a physics degree. Physics is theory. Engineering is the applied (practice) side of that equation. Believe it or not, I have a decent understanding. Its not that I can't make an error or two along the way...

Quote:
There are two theories here in competition: one says that adding extra pixels has an adverse affect on image quality (this is the theory you are subscribing to), the other theory says that it does not.
I think you went a bit far here. I do feel there are trade offs with increasing resolution. Its a pick your poison situation. I try to look at it in pracical terms.

I'll maintain that DR is a trade off. I'll say the volume of data is an additional problem, but its one that Moores Law is correcting as time proceeds. I'm not adverse to the noise arguments except when the sensor hits its gain limits and larger photosites allow more. The D-3 iso 6400 vs the D3x ISO 1600 would be a point in case.

That's the theory I subscribe to.
Quote:
And other the other hand we have statements like:

Quote:
I will not be convinced differently, even with another dozen articles.


I'm still feeling the same way. That camera would be a better camera with half the pixels. Don't let theory get in the way of common sense.

I'm looking at this camera and I find it strange your making a stand here.

I'm willing to go much higher up the chain than this camera, in practical terms. I posted the E-1 pic to show a 5mp image. How much better does it need to be? It, of course, has to do with how its displayed. If printed 30x40, additional resolution might go well. How many 30x40 prints do you do? How close are you to those printswhen you look at them? Its another factor as to if you will receive any benefit from the higher resolution.

It comes down to a question "What/When is enough?" The answer... "It depends." in a practical way.

Greg






fldspringer is offline   Reply With Quote
Old Apr 9, 2009, 6:15 AM   #23
E.T
Senior Member
 
E.T's Avatar
 
Join Date: Jul 2003
Posts: 921
Default

TCav wrote:
Quote:
It seems only Canon has an effective way of dealing with noise: stop it before it starts.
Except that they've overrun advances in sensor technology:
http://www.dpreview.com/reviews/canonsx1is/page8.asp
Really nice looking sky and details.


peripatetic wrote:
Quote:
This is one of the most persistent myths on the internet.

In fact if you take a 12Mp sensor and down-rez the image it will give the same amount of noise as a 6Mp sensor with the same level of underlying technology.
Not in real world because noise of single photosite affects to surrounding pixels:
http://blog.dpreview.com/editorial/2...mpling-to.html
And signal to noise ratio of individual photosite has simply come crashing down to earth and is burrowing to sewer system because of overgross size decrease caused by cramming too many of them into too small space.
E.T 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 4:56 AM.