Comments on: Poll: Megapixels vs. camera sensitivity

If you squeeze more megapixels onto a sensor of a given size, a camera works worse in dim conditions. So what's better, more megapixels or higher sensitivity?

[ralfthedog]

Depth of field control.

You need a full frame sensor to be able to control depth of field. Sometimes you want the foreground and background in focus. Sometimes you want the background out of focus. With a full frame sensor, you can do this by playing with the F-stops. With a tiny little sensor, everything is in focus all the time.

[ImRaptor]

Depth of Field

What are you talking about? You don't require a full frame sensor to control depth of field.
Here is a picture to prove what I am saying http://fc05.deviantart.com/fs22/i/2008/025/0/d/Pulling_by_ImRaptor.jpg

On a smaller sensor you still have F-stop control. I do not need to purchase a 1D for F stop when my Fuji S5200 and my Canon Rebel Xt can both change the f-stop both automatically and manually.

[qubic01]

PixelsSensitivity

Hello,
With better DSPs (and smaller transistor geometries permitting the execution of more sophisticated image-processing algorithms with the same amount of energy) giving cameras better signal-to-noise ratios, sensitivity vs. resolution shouldn't be a tradeoff (at about the time 22-nm geometries come about you might not even need a flash and get around 10 MP equivalent resolution, in particular with the lenses which which give image-focus at ALL distances).
The smaller the transistor geometry, the more sophisticated the treatment to "complete" the details the camera sees (within the same power budget) and the less constraining the tradeoff.

[alegr]

Not quite

Good luck with your 22nm pixels. Never mind that you'll have to have 1:1 aperture to just focus light to 1000 nm spot.
As you go under 2 microns for pixel pitch, you'll have very hard time trying to accumulate at least 10000 electrons per pixel. 10000 electrons means 100:1 signal to noise ratio just because of shot noise, not even taking other sources into account (mostly readout noise).

[kk_junk]

It is better even on same sensitivity

Even on same sensitivity, lower (or reasonable) pixels is better.

http://6mpixel.org/en/

My Fuji 6M superCCD F10 still ROCK!!

[mofo111]

Right On!

I agree 100%. There has been a runaway tendency to equate
pixel count with image quality. Really, now, how many pixels
does one need to make, say, a large 10 inch wide by 6.5 inch
wide picture? Figure on a high quality printer that prints at 300
dpi, and that works out to about 3000 pixels x 2000 pixels, or 6
megapixels. That's all you really need, and that's precisely what I
have on my Nikon D70 SLR. Those cheap 10+ megapixel point-
and-shoot cameras are a joke. All those extra megapixels are
doing is magnifying imperfections in their cheap lenses. In
addition, those extra megapixels increase the file sizes and
increase the noise level per pixel. The only reason 10+
megapixel point-and-shoot cameras exist is because the people
in the marketing department told the people in the engineering
department that they need more megapixels for their ads! Don't
be fooled. You're not going to get better image quality from a
point-and-shoot camera by placing a zillion megapixel sensor
behind a cheap lens.

[markdoiron]

Learn to Take Pictures!

The only reason most folks need anything over 6-MPs or so is because they don't take good pictures! Rule #1, the only numbered rule, for taking better pictures: Get closer to your subject! Those extra MPs are only useful because the snapshooter did not, making cropping necessary. And most folks aren't savvy enough to understand that a cropped photo and a photo of the same area taken close up are two different images--which is why God let man create the telephoto lens!

Bottom line: What digital cameras really need is more dynamic range. It's a crime that this great tech comes with a 1-f/stop penalty over film.

--mark d.
http://www.summitpost.org/user_page.php?user_id=26307

[Phil_Morrison]

Megapixels

I go for MP up to 12-14. That gives about 300 dpi when printing in large format (10" to 12")which is about the limit for typical 35 mm (Kodacolor) reproduciton on a good-decent printer. If I want an outstanding print, then I need about 24MP - but then I need to use a studio with tripod, excellent lens, remote shutter, etc.

It seems that as manufacturers are getting to this limit (12-14MP) they are also increasing the sensitivity. Ideally (in a few years) a full format sensor (35 mm) with 14MP for less than $1000. But take a look at the New Pentax that went to 14MP AND increased sensitivity to 6400! I haven't seen any pictures from it, so I don't know what noise level they have at 6400, but it shows that the two are not incompatable.

[jimsberg]

The Pixel Race

I've been using my 6 MP Nikon D50 for two years (with the upscale kit lens)and making really nice 20" X 30" poster prints! I really don't see the benefit of more pixels, particularly since the results are way better with the D50 than with the 7-8 MP point and shoot cameras I've used and not just because the higher ISO's are better. Sensor size matters!

[shanedr]

Image quality

It's not the consumer that hawks megapixel's, its the industry. Consumers only want to know about the quality of the picture a camera will take. We're still waiting for the industry to inform us instead of misinforming.

[ralfthedog]

How about a better optics option.

Please let us vote for better optics.

[bluvg]

Possible using lower resolution settings?

I don't know if camera manufacturers have tried anything like this, but wouldn't it be possible through some special DSP circuitry to achieve higher sensitivity by shooting a lower resolution image "stretched" across a high resolution sensor? For an extreme example, say you have a 20 megapixel sensor. If you adjusted the camera settings to take a 5 megapixel image, you'd have a four pixel square on the sensor capturing the light for every pixel of the actual picture. Have the DSP do the calculations. Wouldn't that avoid the tradeoff of more megapixels vs. more sensitivity?

[alegr]

It's even better

Most cameras apply different schemes of noise reduction. Usually it consists of edge detection algorithm and adaptive blurring. Areas that are relatively flat, without details are blurred to reduce noise, while eround edges the high-frequency components are preserved.

[skrubol]

Sort of..

Reducing the resolution of an image by any method more sophisticated than nearest neighbor reduces noise. Furthermore, like the other reply mentioned, complex algorithms are already used in-camera to remove noise when not shooting in RAW (NR is generally done on the computer where there's more flexibility when using RAW.) So what you're saying is already done if you use a lower resolution setting on your camera.
The problem is that a 5MP, current technology sensor can put out less noise than a 20MP current tech sensor of the same size after derezing. The reason is that for each pixel in a CMOS or CCD sensor, there is dead area on the sensor that cannot be used for collecting light. The more pixels you have in the same space, the more dead area, which leads to less light captured, and therefore more noise in the end.

[Shankland]

some cameras do this--"pixel binning"

I'm not sure how sophisticated the algorithms are, but a number of compact cameras can combine multiple input pixels into one output pixel through a process sometimes called pixel binning. I've tried it a few times, though, and I'm not generally hugely impressed. Perhaps it's an implementation problem rather than a concept problem, though.

One problem with the approach in SLRs and other high-end cameras, though, is that you can't produce raw images from pixel binning. Or at least it's harder.

[kpp_kpp]

I'd like...

an 8-10 MP camera with an noise free 6,400 ISO that can be
pushed to 12,800 and a true 16 bit dynamic range.

prosumer for $1000

once that is achieve then I'd be happy to jump back on the "uber-
megapixel" wagon.

[skrubol]

Camera with IS or SLR are the only way

If the blur is due to camera shake, and your camera doesn't have image stabilization, you should probably look into getting one with that feature. I think current compact camera IS systems can get 2-3 stops of stabilization (4-8 times the effective shutter speed or iso setting.)
If it's because your subjects are moving, or your camera already has IS, I wouldn't hold your breath on compact cameras with better sensitivity. Sensors are getting relatively close to theoretical maximums in sensitivity, and cameras are going to continue to gain megapixels more than they gain sensitivity. The only real way for better sensitivity is more sensor.

[btljooz]

This is why......

I :x [b]LOVE[/b] :X my 28 year old Canon AE-1!!!

Won't stop using it until I can't get film for it!!! B-)

[btljooz]

OPPS!

First, :X should be :x.

Second, my memory crashed for a sec and I my math was corrupt.

My AE-1 is 25 years old. I bought it when they FIRST came out in 1983.

See it here:

http://en.wikipedia.org/wiki/Canon_AE-1

[ElMartino1]

Look at high ISO photos from a Canon 5D (link inside)

to see how Full Frame, even with a DigicII processor that's over 2 years old performs.

All of the photos in this FlickR pool are ISO 800 or higher.
http://flickr.com/groups/canon5d/discuss/72157600052412462/

I *adore* my 5D, and I only just got it. The 6MP 10D and 300D can't even get beyond ISO 400 without looking like someone poured red pepper seasoning salt all over the picture.

[Shankland]

Thanks, that's a great gallery

I like these shots--perhaps because the difficulty of low-light shooting makes that sort of photo rarer and therefore fresher.

[ImRaptor]

Will accept xD donations

I will gladly accept a 5D or 1D as a donation. No argument from me on the quality improvement over my Canon SLR.
The best I can justsify right now is to replace my Xt with a 40D for now.

The gallery is a great example of what is gained with a more expensive sensor and why the better quality in low light trumps megapixels.

[ralfthedog]

Comment to the author.

Please admit that this story was just an excuse to show off the cool shot of the car tire.

:)

PS. I can get the same shot with a slow shutter speed. I just need to add some noise afterwards.

:) :)

[Shankland]

not quite so easy

That car was going by at 80+ mph, which means the top rim of the wheel was moving at 160mph, which means a slow shutter speed would have been not quite the same. I could show you some of the blurry ISO 200 shots if you want. :)

[skrubol]

You are confused

Just because you use an EF-S lens doesn't mean your crop factor goes away. 50mm is 50mm. One of the EF-S lenses that goes to 17mm at the wide end will have pretty much the exact FoV as say the 17-40L EF lens when put on an APS-C camera.
Most of the EF-S lenses focal lengths are meant as equivalents to common EF lenses. The 18-55 is approximately equivalent to 28-85, a very common starter lens. The 17-85IS is equivalent to the 28-135IS, and the 10-22 is equivalent to the 16-35L.
Also, due to the different focal lengths, the DoF of a 30mm f/1.4 will not be as narrow as that on a 50mm f/1.4. It won't even be as narrow as 50mm at f/1.8. That is why smaller sensors cannot get as narrow DoF as larger sensors.

I meant to post this with my response yesterday. This is the calculator I've used to do my DoF math. Play for it a bit and maybe you'll understand the relationship between sensor size, focal length and DoF.
http://www.dofmaster.com/dofjs.html

[skrubol]

Oops, meant to post as a reply to a comment above

NT

[georgesvh]

Dynamic range

Neither matters, as current camera technology is an inaccurate
technology by a far stretch. The real issue is in cameras is the
fairly flat Dynamic Range. Simply put, Dynamic Range is the
sensitivity by which receptors can still see detail in the mid-
tones while their is lots of bright light, or lots of darkness.
Dynamic Range is a function derived from many factors
(including pixel density) but the Dynamic Range of your eyes is
still much, much better than any camera can produce. We should
invent cameras that measures light before appropriately before
it hits the sensor, just like our eyes do. Till then, the
combination of pixel density and exposure metering is probably
the most important factor.