The Truth About Digital Cameras

[DavidTO]

At least, according to David Pogue (http://pogue.blogs.nytimes.com/2006/11/21/21pogues-posts-2/).


Discuss.

[Tommyboy]

I admit that the megapixel thing has always confounded me.

Say what you will, I made outstanding 8X10 prints from a Canon 2.0 megapixel camera. Not "what I think was excellent," but really sharp, clear, colorful, fine-grained prints.

Seems Pogue's onto something. . . .

[Mongrel]

if this gets out....

the whole *industry* is doomed!

:cry

[Harryb]

It isi the truth basically. I could frame shots from my 5 mp Sony 717, my 4 mp D2H, my 6 mp D100, my 10 MP D200 and my 12 MP D2X and from a normal viewing distance you would be hard pressed to say which camera took which shot.

Now if you viewed them at 100% on your PC screen some differences would start to show.

The biggest difference I've found is in my ability to crop shots. Thats when I can really see the difference between my old D2H and my D2X.

However when viewing shots in normal circumstances there really isn't much difference to be found.

[wxwax]

At least, according to David Pogue (http://pogue.blogs.nytimes.com/2006/11/21/21pogues-posts-2/).


Cool story. Supports dgrin saying you only need tosave at PS 10, not 12.

Discuss.

Have I ever mentioned how obnoxious this is?

[DavidTO]

Cool story. Supports dgrin saying you only need tosave at PS 10, not 12.



Have I ever mentioned how obnoxious this is?


:lol3 I just wanted to be like Andy.


And yes, I know what you mean. Feel free to edit. :D

[pathfinder]

He did not specify what cameras were used though.

I, for one, do not believe there is no difference between a 5 Mpxl Canon G-5 ( which I own) and a 12Mpxl Canon 5D ( which I also own)

Even when shot in RAW, the Canon G-5 images cannot compete with those of a DSLR, let alone the 5D.

The image sensor size and quality still counts, as does the size of the pixel wells themselves.

IF he is only referring to point and shoots, with their smaller sensor size, then, I am inclined to agree that higher pixel counts do not make much difference - they may even make a camera have more noise in the images.

[bham]

From what I read it sounded like he took one 13 mp image and then created a 8 mp image and a 5mp image from that original 13mp image. And then had the lab print all 3 at the same 16x24 size. Now the prints are gonna be hard to tell apart mainly because the lab will up rez the images as needed to produce the correct dpi for the print size. Just as EZ Prints does for our prints.

From this example I can see where the difference would be hard. Now put those three files on a large screen at 100% and I think the difference would be much easier to identify.

He didn't use 3 different cameras to capture 3 similar images. Its not all in the megapixel but also in the quality of those megapixel.

You take a older DSLR body vs a new point & shoot of the same megapixel and I would be pretty sure the DSLR would produce a superior image. Probably a better sensor, and better post processing done by the camera. I think some of the technology from the earlier version of the DSLR has probably found its way into the point & shoots but if I was a camera manufacturer and realized that the turnover in cameras is every few years and that most people who buy the point & shoots hardly can figure out the camera, let alone research the ins & outs of the guts of a camera, then this is probably the place you could scimp and people not realize the difference. People have been trained to think, more megapixel better camera, so the industry caters to that and so does their marketing. If they have a budget for a camera and it comes down to a great image processor ie Digic III or a bigger sensor, I know what is going to win that battle every time. So instead of a 8mp P&S with a great processor, you get a 12mp P&S with a Digic 1.5. (Not exactly but you get my point)

[pathfinder]

Like Canon omitted the RAW capability from their new PRO Shot G-7, and has received real criticism for it. The G-7 looked like a really neat camera, but without RAW capability, my interest lagged quickly.

[ziggy53]

I suggest that number of pixels does matter, depending upon:

Subject matter. (Some subjects don't need as much detail to "look right".)

Method of printing. (Matters greatly. Some printing methods overlap while others are more discreet and defined.)

Method of presentation and/or display. (Matters greatly. Image projection appears to be the least tolerant of low resolution. Line-doublers and image interpolation are common methods of tricking video consumers into thinking they are seeing image detail not really there. Faroudja Labs has done amazing work on this.)

Degree of "lossy" compression used. (More resolution tolerates more compression with less visible impact. DCT based compressors, JPG for instance, use discreet "cells" and if those cells are relatively large, with respect to overall image size, the image suffers more rapidly at higher compression ratios. Wavelet and curvelet compressors show significant improvements for lower resolution images and higher compressions.)

Amount of intrinsic noise. (Noise is treated by our vision and brain as detail. The more noise, and especially the larger the visible grain of the noise, the less important is the actual image detail, so the noise "masks" the significance of the image detail. NASA did a major study on this and developed some interesting results.)

I submit that image resolution is not a simple subject and that there will be continued "discussion" for years to come.

ziggy53

[greenpea]

This makes me curious as to how far people have pushed their megapixels. In other words whats your largest print size to megapixel ratio.

I did a 20 x 30 print using a 6MP image from my D70, it looks great!

The article makes me wonder how large of a print I could make and still retain high image quality, using my first (and still operational) digital camera, a 1.5 MP Kodak DC260 (http://www.dpreview.com/reviews/specs/Kodak/kodak_dc260.asp).

[mr peas]

It isi the truth basically. I could frame shots from my 5 mp Sony 717, my 4 mp D2H, my 6 mp D100, my 10 MP D200 and my 12 MP D2X and from a normal viewing distance you would be hard pressed to say which camera took which shot.

Now if you viewed them at 100% on your PC screen some differences would start to show.

The biggest difference I've found is in my ability to crop shots. Thats when I can really see the difference between my old D2H and my D2X.

However when viewing shots in normal circumstances there really isn't much difference to be found.

Yeah I agree. My Sony 828 looks great at prints during daylight, but so does my Canon XT...both being 8mp. But when it comes to low-light no-flash shooting, you can tell the Canon from the Sony from the noise each produces.

But if you need to crop something out, that higher MP camera will be your best friend lol. The REAL test would be, a cropped portion print of an image from three diff. MP cameras. That would discern the thin red-line. :wink

[claudermilk]

This makes me curious as to how far people have pushed their megapixels. In other words whats your largest print size to megapixel ratio.

I did a 20 x 30 print using a 6MP image from my D70, it looks great!

The article makes me wonder how large of a print I could make and still retain high image quality, using my first (and still operational) digital camera, a 1.5 MP Kodak DC260 (http://www.dpreview.com/reviews/specs/Kodak/kodak_dc260.asp).

I've done a 16x20 off a 3.3MP Nikon P&S. Up close you can see the beginnings of pixellation, at normal viewing distance it looks great. The sensor size is just one factor in a complex system, nice to see an article that points towards that reality.

[gus]

Cool story. Supports dgrin saying you only need tosave at PS 10, not 12.



Have I ever mentioned how obnoxious this is?
Not wishing to pick on david but i couldnt agree more ...twice. Its just one of those little things that urks me also...in particular when a waiter places food on your table & says 'enjoy'...urghhh !

[DavidTO]

Not wishing to pick on david but i couldnt agree more ...twice. Its just one of those little things that urks me also...in particular when a waiter places food on your table & says 'enjoy'...urghhh !


It's funny....I was really just doing my Andy impression....

[ian408]

I want to see his test repeated without resizing. After all, his work (IMHO)
does not compare apples to apples. It simply states that when an image is
resized and printed, it is difficult to determine which was which.

If he used a 5 & 13 mp cameras from Canon or Nikon, shot RAW (to
eliminate in camera processing), then did his printing without doing
any sort of resize, I'd be more convinced. If he really wanted to, he
could do a studio shoot to eliminate any "lighting" issues.

Mind you I'm not suggesting he's right or wrong just that his testing
methodology is questionable.

[gluwater]

At my old job I had access to an HP Designjet 5000 (http://h10010.www1.hp.com/wwpc/us/en/sm/WF10a/18972-236251-236266-12600-236266-25300.html). I wanted to see how large I could print from my Olympus 750UZ, a 4 MP 10x optical zoom P/S. The largest I tried was 36x48 and I was very pleased with the results. The thing I found most important was how I set the printers RIP, not how many MP I had. I think with todays printing technology and advanced RIPs you can really print whatever size you want from most cameras, regardless of MP. But that's just my .02.

[ziggy53]

It's funny....I was really just doing my Andy impression....

Better keep your day job! :rofl

(Observe the humorous icon which removes all doubt from the humorous intent of the comment. Nyuck, nyuck, nyuck, woo, woo, woo woo woo woo woo.)

Happy Thanksgiving everybody,

ziggy53

[ziggy53]

I want to see his test repeated without resizing. After all, his work (IMHO) does not compare apples to apples. It simply states that when an image is resized and printed, it is difficult to determine which was which. ...

It may simply be that his method, process and procedure obscure the difference. In other words, he may have chosen a system which is not sensitive beyond a certain resolution.

I assure you that there are many systems, and many applications/uses, which "are" sensitive to resolution and total pixel count.

Hasty generalizations for the purpose of simplification are often flawed.

ziggy53

[El Kiwi]

Still, if the rumours are true and Canon will be replacing the 5D early next year with a "low end" 8MP full frame, and a "high end" 16MP one, I know where my money is going. You'd have to print those enormous ads they have on the sides of buildings to tell the difference.

[Andy]

This makes me curious as to how far people have pushed their megapixels

Here's 8 megapixels, small sensor, Sony 828 printed at 6ft x 4ft. :thumb

http://www.moonriverphotography.com/photos/12718508-S.jpg

[DoctorIt]

This article brings up the question I'm constantly asking: why doesn't anyone concentrate on dynamic range and pixel quality rather than pure number of pixels? I know damn well that my 4mp D2h took better images than my 5mp Pentax Optio. So clearly, its not just quantity, its quality - something I'd like to understand better.

And an edit: my current dream camera, seems there is at least one company thinking in quality - Fuji S5 (http://www.dpreview.com/news/0609/06092502fujifilms5pro.asp), a D200 with a badass 6mp sensor :nod

[JCDoss]

And an edit: my current dream camera, seems there is at least one company thinking in quality - Fuji S5 (http://www.dpreview.com/news/0609/06092502fujifilms5pro.asp), a D200 with a badass 6mp sensor :nod

But the S5 will have pretty much the same sensor as the S3. The difference will be in the handling. I was excited about the S5 for a while, but I don't want to wait till Feb... I will pick up a D200 soon.

But to stay on topic, I agree that the test seemed to lay to rest the myth of "more is better," but totally failed to address the quality side.

[Khaos]

This article brings up the question I'm constantly asking: why doesn't anyone concentrate on dynamic range and pixel quality rather than pure number of pixels? I know damn well that my 4mp D2h took better images than my 5mp Pentax Optio. So clearly, its not just quantity, its quality - something I'd like to understand better.

And an edit: my current dream camera, seems there is at least one company thinking in quality - Fuji S5 (http://www.dpreview.com/news/0609/06092502fujifilms5pro.asp), a D200 with a badass 6mp sensor :nod

Increased Dynamic Range is the holy grail. That's why the human eye is so amazing.

I think it's a matter of economics and stupid consumers. People hear bigger numbers and automaticaly think its better. In that theory, the 10 mp 400D should be better the the 8 mp 30D, which we knoe isn't true. Having to spend a tremendous amount of R&D dollars on increasing dynamic range isn't worth it when the only people who care are small in number, when increasing mp size means a lot to your average consumer.

[wxwax]

Increased Dynamic Range is the holy grail.

It is for me. I sometimes wonder if it is for camera makers.

[Awais Yaqub]

dont know about 16x24
but i printed 10x8 with standard size vga there was no distortion and people loved it . and i did a lot prints for relatives 10x15 @ 1 Mp no one said anything negative :D
Wont there be huge differences when we go to even big printing i mean if Andy got similer photo at 1 mp and printed that large :huh

[DoctorIt]

Increased Dynamic Range is the holy grail. That's why the human eye is so amazing.

I think it's a matter of economics and stupid consumers. People hear bigger numbers and automaticaly think its better. In that theory, the 10 mp 400D should be better the the 8 mp 30D, which we knoe isn't true. Having to spend a tremendous amount of R&D dollars on increasing dynamic range isn't worth it when the only people who care are small in number, when increasing mp size means a lot to your average consumer.Yup, you're right, it's all about the benjamins!

[thebigsky]

I'd like to see them blow up an image taken with my Olympus Camedia C-820L and still claim resolution doesn't matter.

Charlie

[jimf]

There's been some pretty good discussion here, so what the heck I'll toss in a few cents.

It really ought to be obvious that the biggest limitation on most P&S cameras is not the sensor, but the lens. It's my experience that even the best P&S cameras see little benefit from sensors with resolutions greater than about 3mpel, and ofter suffer worse sensor noise artifacts at higher resolutions. It's also been my experience that 3mpel is fine for an 8x10 print; not tack sharp, but very usable. Thus I tend to tell people that it's not worth their money to buy a P&S with greater than 3mpel resolution; if you need better quality than that, you need to buy a DSLR to get better lenses.

The biggest image impact I have seen is not the result of resolution but of compression. One of the first tests I did when I bought my 300D was to shoot a RAW frame, convert it to TIFF and JPEG at high quality in Photoshop, and had a local pro shop print both at 5x7. Then I went around showing the two prints to random people to see if they could tell the difference.

Before I did the experiment I didn't think they'd be able to, not with a 6mpel source printed at 5x7. But when I got the prints back I could immediately tell which was which, and so could almost everyone else.

I showed more than 20 people at home and work those prints and asked them to pick the best print. All but one person picked the TIFF print, and the one who didn't has macular degeneration (ie very poor eyesight) and couldn't see any difference at all. The odds of that happening randomly are more than one in five hundred thousand; pretty definitive.

My conclusion was that JPEG compression significantly degrades image quality. In later tests, I would find that shooting RAW at lower resolution yielded superior prints to shooting JPEG at high resolution. The difference is not dramatic, but it's there. (I note that my reason for using RAW is for greater exposure lattitude, not improved resolution, but it's nice to have both.)

Another thing that has become very obvious over the years is that the quality of JPEG compressors varies quite significantly. Some of the P&S cameras out there have such lousy compressors that their resolution is almost immaterial. For an example of a really bad one, look at the Olympus 720sw. That's nominally a 7mpel camera, but it produces rather worse images than the Canon Powershot S330 (2.0mpel) in both color and clarity. (The kinds of artifacts I'm seeing appear to be related to the compressor rather than the sensor; I would bet almost anything that they have "optimized" the JPEG compressor for improved performance.)

In short, I'm one of the people in the camp that says that you don't need high mpel numbers in most cases. I too find that higher resolutions are most useful in cases where you need to crop. If I were to pick attributes of a camera that are worth paying money for they would be focus and recycle times.

[Khaos]

Nice post.

That's probably why magazines that have switched to digital require submissions be in tiff and not jpeg.

[rutt]

One of the coolest sessions of the computational photography course (http://groups.csail.mit.edu/graphics/classes/CompPhoto06/syllabus.shtml) I audited last spring was on "future cameras (http://groups.csail.mit.edu/graphics/classes/CompPhoto06/html/lecturenotes/futureMay112006.pdf)". The slides are a little hard to understand out of context, but better dynamic range consumed a lot of the lecture. Oh, and don't miss this (http://groups.csail.mit.edu/graphics/classes/CompPhoto06/html/lecturenotes/08_HDR.pdf). This kind of academic advanced thinking is likely to have a big impact on products in the long term.

Just to stretch your mind, here is the real holly grail of future cameras: don't have to set exposure because dynamic range is great enough to capture the sun and the unlit interior through a window at the same time. You don't have to focus because the sensor captures all the info to make focus and DOF decisions in post. The thinking is that Moore's law applies to sensor density and we already have enough resolution, so what should we do with all that area?

[JCDoss]

...don't have to set exposure... You don't have to focus...

OK, but do you still have to go to Grandma's birthday party, or will the camera take itself? :rofl

[DoctorIt]

You don't have to focus because the sensor captures all the info to make focus and DOF decisions in post. The thinking is that Moore's law applies to sensor density and we already have enough resolution, so what should we do with all that area?I remember you telling me about this. Really crazy stuff, but as you say, shouldn't really be science fiction, Moore's law should hold in digicams as well.

[rutt]

HDR sensors that don't need exposure setting at shoot time and produce images which can be focused in post may seem like science fiction. But we are already past one milestone on the road to capturing more information at shoot time than we need for the final shot. Consider B&W. I contend that the very best way to shoot B&W today is to shoot digital color and convert in post. The great B&W film masters used all kinds color filters when they shot in order to enhance the contrast that mattered to them. We can delay this decision and even selectively blend the different color channels in post. This allows us, for example, to use the green channel for faces while using the red for background skies (just start with the green and apply image, 100% darken to add that good red channel sky without touching that good green channel face, flatten and repeat to intensify the effect.)

If you've seen Good Night and Good Luck, you've seen how effective this can be. This film was shot in digital HD color and converted to B&W in post. For my money, it's the very best looking B&W film of all time.

Ansel Adams would be in heaven with all the options we already have available to us. And just wait.

[El Kiwi]

...here is the real holly grail of future cameras: don't have to set exposure because dynamic range is great enough to capture the sun and the unlit interior through a window at the same time. You don't have to focus because the sensor captures all the info to make focus and DOF decisions in post.

See, while I can see why this is appealing, and I'm sure it's only a matter of time, photography sure will be boring when it happens.

[rutt]

See, while I can see why this is appealing, and I'm sure it's only a matter of time, photography sure will be boring when it happens.

Well, not exactly. It will just move more of the entertaining part from shoot to post time. Suppose you want a print on paper of the beautiful girl with the sun in the background. OK, you have all the bits, but paper doesn't have the gamut. Now what? Suppose you want to deemphasize something?

As I said, it will be like B&W. Conversion at post time can accomplish all that can be accomplished with papers and different kinds of film at shoot time. And more.

Much more powerful tools open the door to all kinds of possibilities we haven't imagined yet.

[Leethal]

I think that Pogue's test was destined to prove just what he wanted it to prove. Maybe he just wants to stir up a little traffic on his site.
Let me explain,
" One print had 13-megapixel resolution; one had 8; the third had 5. Same exact photo, down-rezzed twice, all three printed at the same poster size."
Look at his second sentence.... Does it mean that he down-ressed the 13MP and the 8MP down to 5MP before he printed all three at "poster size"?
If so, then all the extra details in the larger 2 images were LOST when he re-sized them DOWN TO 5MP! He theoretically started out with 3 images that were 5MP each! Of course they are going to look identical!!!
This is just my opinion, of course. Maybe I misread his "testing" procedure!
If this is the case-my apologies are in order!
Oh yeah - discuss this amongst yourselves :D

[Khaos]

See, while I can see why this is appealing, and I'm sure it's only a matter of time, photography sure will be boring when it happens.

Not at all. When you eliminate obstacles you also open up doors for more creativity. Everyone will take technically perfect exposed shots, but will they now how to play with light, DOF, and color to get those special shots? Nope. Making something easier really doesn't take away anything, it just makes you able to do more.

[ian408]

I think that Pogue's test was destined to prove just what he wanted it to prove. Maybe he just wants to stir up a little traffic on his site.


I think a number of people have called him on this. I haven't read his
response yet.

John, I think you've made two very important points. The second is about
more powerful tools being enablers and first, and perhaps most important,
is what we will do with the area (Moore's Law).

And I would agree with you on Adam's glee in today's technology.

[DoctorIt]

And I would agree with you on Adam's glee in today's technology.Wouldn't it be neat to see what some of the most famous masters of the past would do with modern technology? As I recall, Adams was a pioneer in developing/printing technology back then, no?


Great thread by the way, I've gone ahead and given it some stars :thumb

[rutt]

Imagine Man Ray unleashed with Photoshop! Actually I don't think I can.

Wouldn't it be neat to see what some of the most famous masters of the past would do with modern technology? As I recall, Adams was a pioneer in developing/printing technology back then, no?


Great thread by the way, I've gone ahead and given it some stars :thumb

[MarkR]

I think that Pogue's test was destined to prove just what he wanted it to prove. Maybe he just wants to stir up a little traffic on his site.
Let me explain,
" One print had 13-megapixel resolution; one had 8; the third had 5. Same exact photo, down-rezzed twice, all three printed at the same poster size."
Look at his second sentence.... Does it mean that he down-ressed the 13MP and the 8MP down to 5MP before he printed all three at "poster size"?
If so, then all the extra details in the larger 2 images were LOST when he re-sized them DOWN TO 5MP! He theoretically started out with 3 images that were 5MP each! Of course they are going to look identical!!!
This is just my opinion, of course. Maybe I misread his "testing" procedure!
If this is the case-my apologies are in order!
Oh yeah - discuss this amongst yourselves :D
The impression I got was that he started with a 13 mp image printed out. He then down-rezzed it to 8 mp, printed that copy, then down-rezzed it again to 5 mp and printed that copy. He then had three images: 13 mp, 8 mp, 5 mp, and asked viewers to compare them.

What really amazes me is how ... angry ... so many of the comments to Mr. Pogue's post have been. You'd think he'd written about the joy of clubbing baby seals, or how dolphin meat adds flavor to canned tuna.

I think that marketers have spent so much time selling megapixels, that consumers have become brainwashed into thinking that it's the most important measure-- or maybe the only measure-- of a digital camera.

[ian408]

Wouldn't it be neat to see what some of the most famous masters of the past would do with modern technology? As I recall, Adams was a pioneer in developing/printing technology back then, no?


Absolutely. If you look at prints Adams made, especially when separated by
time, you may notice they're processed differently.

I'd like to think he would embrace the changes in photography since his time.

[sgbotsford]

At least, according to David Pogue (http://pogue.blogs.nytimes.com/2006/11/21/21pogues-posts-2/).


Discuss.

I question that 13 Mpx rating. Until the camera companies start putting out larger sensor chips in their cameras, you are collecting in what, about 2/3 of a standard 35 mm frame. Call it 20x30 mm. A really good lens resolves around 100 line pairs per mm. So there is only 2000 x 3000 line pairs of data. (Yes, this is an over simplification. Beat me with a stick...:rofl ) So if you had a sensor that was nominally 3000 x 5000 pixels you have more than one pixel sampling each effective resolution unit of the lens. And remember that 100 lp/mm is under optimum conditions. Use a zoom; use a filter with no longer perfect coatings, have a finger print or two, and getting down to 50 lp/mm comes all too easily. At that res, you are sampling each unit of lens resolution with a 3x3 block of pixels. -- 1/9 of what you started with.

I'm not surprised that you can de-res an image down to 5 Mpx and get prints that are difficult to tell apart.

It's been years since I've followed the tech specs of lenses. I may be wrong about some of the numbers here, but I think I'm in the right ball park.

I do know that a lot of filters work better with more pixels. It's easier to avoid halos when sharpening a high res image. Some company claims wonderful results with a set of PS filters that work by increasing the resolution, doing their magic, then decreasing it again. (Can we say CPU intensive...)

[erich6]

As Ziggy mentioned pixels DO matter but only in relation to other aspects of the entire photographic process. At the camera, it depends on what kind of lens you have in front of the sensor.

In practice, the types of lenses in compact cameras end up being the limiting factor and therefore you don't get much benefit from increased pixel count. It's a different story with SLR cameras.

Erich