Megapixels – Are they important?
 

Popular opinion states that the more megapixels a cameras sensor has then the clearer the resulting image and therefore better quality. Camera manufacturers have for some time been involved in a race to create affordable high resolution image sensors within the bounds of AP-S and 35mm format. This article is concerned only with cameras from manufacturer Nikon and assesses resolution from three models D2Hs, D70 and D2X which have native resolution of 4.1, 6 and 12.4 megapixel sensors. Whilst the author does not hold an exhaustive range of lens charts, images are examined using the same test piece for a real life comparison. All images are assessed on equal terms but due to logistical issues concerned with supplying every reader of this article with a printed 6 x 4 photograph images are presented at monitor resolution - 72 dpi (Dots Per Inch) therefore every reader can view a representation of each photograph suitable to computer monitors.

A megapixel is a one Million picture elements so the understanding that more equates to better images is a logical one. However the truth is often realised when a photographer purchases a new higher pixel count camera and discovers that the images are often worse then their older lesser pixel count camera. When the same AF-S area of approx 24 x 16mm is used the only physical way to arrange more pixels is to employ smaller and smaller photo-sites. This may result in increased noise (unwanted speckles due to non-linear output of each pixel when subject to equal light energy) as a smaller element is not exposed to the same amount of light energy or photons as a larger one therefore electronic amplification is required to increase a pixels output level which is governed by aperture, exposure time and ISO speed. In theory all cameras regardless of manufacturer should output the same brightness for a given ISO, exposure and lens aperture. If noise prone electronic amplification is not used then another way of keeping within these bounds is to reduce the ISO speed rating i.e if the sensor output is low due to an increased pixel count then the output could be rated at different ISO level i.e. 1/125 at f4 produced half the level it should at ISO200 then a manufacturer could claim the sensor output is at ISO100 whereby the photographer can increase ISO speed to increase output. A higher pixel count sensor as well as recording more detail will record more flaws such as lens aberrations and poor technique from the human photographer. If a chromatic aberration exists from a consumer lens used on a 4MPx (MegaPixel) camera then the extent of colour shift maybe across ½ pixel and not recorded but the same lens used on a 12MPx camera could have colour shift across three pixels each side of a high contrast edge resulting in six pixels being exposed to lens aberration which could definitely be recorded. The same applies for poor technique an edge may be blurred across one pixel on a 4Mpx but this may be three with a 12Mpx image to show perfect results viewed on computer monitors at full pixel count from higher resolution sensors perfect technique must be employed.


Camera sensor specifications

D2Hs, 4.1 Million Pixels (2464 x 1632) 23.3 x 15.5mm. JFET LBCAST
D70, 6.01 Million Pixels (3008 x 2000) 23.7 x 15.6mm. CCD
D2X, 12.4 Million Pixels (4288 x 2848) 23.7 x 15.7mm. CMOS

Notice differences in physical size between all cameras and difference in sensor technology as the motivation of this article is image quality between pixel count this will be ignored and has little relevance when images are printed.

Full specifications can be found below
D2H (as only a preview D2Hs) can be found
http://www.dpreview.com/reviews/nikond2h/

D70
http://www.dpreview.com/reviews/nikond70/

D2X
http://www.dpreview.com/reviews/NikonD2X/


The Test
Testing each camera is not about any specific camera, its features or quality instead the test is concerned only with megapixel output so differences in colour, exposure and noise are ALL ignored. It must be reiterated that this test is purely concerned with megapixels.

A teddy bear was selected for fine detail in its synthetic fur and especially a course weave ribbon around its neck. All images recorded using the same 50mm f1.8 lens at f8, reviewed by Bjørn Rørslett ( http://www.naturfotograf.com/index2.html ) 'having no noticeable difference between f4 and f8'. Aperture f8 was chosen at the end of this range to provide maximum depth. A solid slik professional tripod was used to avoid any motion shake and all recorded using Nikon Camera Control Pro (computer tethered) to avoid any human induced vibration.

Fine Detail
An single area of course weave ribbon was chosen as the focus area The examined area is a 500 x 500 pixel crop and because of the difference in pixel count this area will show a larger or smaller patch of the photographic subject.



4.1Mpx file detail crop
Minimal detail is shown at native resolution but is this acceptable?
http://www.foxaviationphotography.co...mg0014crop.jpg


6.01 Mpx fine detail crop
More detail in the weave than 4.1Mpx sensor is this acceptable?
http://www.foxaviationphotography.co...mg0018crop.jpg

12.4 Mpx fine detail crop
Lots of detail in the fine weave much more than 4.1 and 6.01 Mpx sensors and quite a surprise.

http://www.foxaviationphotography.co...mg0016crop.jpg


All three images are acceptable but the level of detail provided by the 12.4Mpx sensor surpasses any expectation based on the former two. Although when looking at the fur in each example there seems to be little difference. Therefore lower resolution sensors are perfectly acceptable with certain scenes and a higher pixel count sometimes does not provide better sharper images within that scene.

Printed representations

All three images have been resized using Nikon Capture NX and their colour profile converted to sRGB for Internet viewing. Monitor resolution of 72dpi would require an image of 432 x 288 to represent a 6 x 4 photograph which is very small and would provide little value. Printed representations are instead 800 pixels high to cover most readers monitor and would equate to a photograph being nearly 11 inches along the vertical side.

http://www.foxaviationphotography.co...mg0014full.jpg

__________

http://www.foxaviationphotography.co...mg0018full.jpg
__________

http://www.foxaviationphotography.co...mg0016full.jpg

Is it possible to judge which is which? - NO!

Therefore anyone wishing to share photographs using the Internet can do so using the most modest of photographic equipment without being bogged down by the complications of megapixels. Printed photographs to approx 6 x 4 at 300dpi can be produced without any real difference.
Extra megapixels are required when printing large images but some claim the advances in bi-cubic resizing permits photographers to produce high quality enlargements from 4Mpx sensors although as seen in the native 500 x 500 crop examples that more megapixels does provide more detail if used without resizing. Extra megapixels are require more memory card and hard disk space, if using a camera to its optimum RAW should be used necessitating a very powerful computer to process and hold in memory a highly complex and large RAW image. In the examples RAW image size were in a range of 3.5 to 10.3MB hence a linear relationship between megapixel count and storage requirement although based on experience a non-linear relationship exists between computer performance and RAW image size, requiring more than three times the power to process a 10.3MB RAW file when compared to 3.5MB output from a 4.1Mpx camera in the same period of time.

So, are megapixels important?
In summary, NO! Megapixels do offer greater scope to select an area of a photograph to crop and provide greater fine detail but often negated when printing standard size photographs and sharing using the Internet. More megapixels require more hard disk capacity to hold for long term storage. Greater processing power is also required if shooting RAW mode and often have greater amount of noise. Lower megapixel sensors gather more light resulting in cleaner images but with less detail but require less storage capacity.

Stephen,

I think you've answered your own question here. If you're doing selective crops of images that you then print at A3, then yes, pixels count. If you're taking pictures for viewing on a screen, or printing 6*4, then you can get away with just about anything, and the higher pixel count will be meaningless.

Besides, there's some photographers with some very high MPixel count cameras that still take some bad photos. Likewise, you seem to do ok (tongue planted firmly in cheek here) with your D2. Those who look at the pixels miss the big picture.

Duncan

How many times has another photographer walked up to you and asked... "So, How many Megapixels is that then?"

It happens to me almost every time I go to an airshow!

Perfect set of articles, Stephen, that should slam the lid on the megapixel argument. I'd advise anyone else still in doubt to look at some of Mose Peterson's wildlife photos that he took with an early Nikon - 2.8Mp - they are just outstanding. It's the photographer's knowledge of the subject, light, and composition that makes the image, not the number of pixels.

For internet I have experience with the D2X. Images with a high level of detail I have had to shrink so much they just look nothing like the origional. Some are then not worth posting even with WPF's generous file size allowance.

Processing power - I had to get a pc upgrade before I could handle my RAW files and have just invested in a large capacity stand alone drive to hold the tiffs.

Can I see the difference on screen between a D2X and a D100 image ........ well yes I can, but then a lot of my pics are for a better word ' hard ' subjects with a lot of detail. Soft subjects like flowers, then they follow Stephens examples above and you would be harder pressed to tell the difference between D100 and D2X on resized to 1024 x 760 images. As for 800 x 600 then even in a detailed image so much is lost in the shrink that a 6mp could look better than a 12mp.

The one other thing that will make a significant difference is lens quality. Top lens on a 4mp will probably give a better resulting image than mediocre lens on 6 - 8mp camera. Foxy probably does not have access to such to prove me wrong. :D

Don

50mm f1.8 was the lens of choice as resolution is high and could easily provide good results on both 4 and 12 Mpx cameras.

It seems there are applications for both low and high pixel count sensor cameras.

At what point does digital noise raise its head? For small sensors as found in compacts a lot of megapixels seems counter productive as the noise suppression algorithms needed seem to make the photos worse rather than better. The 6mp compact I use a lot of the time seems about the right compromise at the moment. SLRs with their bigger sensors will have a higher ceiling.

Excellent article, a question I have wanted to ask ....
Something I note is that when stored in the camera,
a 9500Mp picture takes up 2.2 Mb. If I rotate the photo
it become 1.3mb; and if I alter the contrast, or something,
and save again, it becomes 800Kb. What is being lost,
if anything?

treeve,

All sounds a bit odd to me. I suggest you post this and a pic in the Digital darkroom forum with details of your imaging program. People could then use that pic, and go through your routines and detail their results.

Don

I agree. I think the mp race has gone a bit mad. I have just looked through DPR but have not found a test of a 10mp compact. I know they have commented on this subject though.

Don

Totally understand the choice for this one Stephen. But I am sure lens quality will have an impact.

Perhaps I should have a go.

D2X plus Stevies £200 Sigma 28-200 compared to the D100 and 55 micro or 105 f2.5.

Don

When you save as a 'JPEG' (jpg) file a certain amount of detail in the image is thrown-out, so to speak, to keep the file size small, but is then 're-invented' when you open up the image again.

Some of the finer detail gets lost and, though it may still look fine as a photo, what you've got is a sort of 'simpler' image so every time you alter anything you have to 'save' when you close the image down and by saving as a JPEG again you throw out detail from the 'simpler' image.

This is why the file size is now smaller - the simpler the image the more the system can throw out because it's easier to re-invent it back again (check the file-size for a jpg image containing a large area of one colour).

Doing it several times will just make the image worse and worse each time as more and more fine detail is lost so always save as an un-compressed file type such as 'TIFF' (tif)

Adey is dead right and I missed this critical bit when I read the post. Depending on the program you are using you will have various levels of compression. This suggests that your program is using medium compression levels with each save. Try a pic where you perform all operations and only do a save at the end and compare final file size with one that has multiple saves. Also check if you can adjust the compression level in your ' save as '.

Paint Shop Pro 7 & 8 work like that. If ' save as ' is set at 10% compression then that setting is also applied to ' save '.

Don

Thank you very much for the advice Adey and Don ...
I had not noticed that part of the Paint Shop Pro 7 "save" system.
I keep all my digital jpeg pics as taken on camera, copied on to
two external drives and two DVD copies. Always have four originals
and then make adjustments on the PC and laptop to a secondary copy.
So nothing lost, as it were. A bit of tinkering under the bonnet of PSP.


Adey, you have raised another question for me, now.
My camera is set to record jpegs, which can produce 4.4Mb pics
on Fine setting, or 2.2 Mb on Normal. If I was to copy these across
from the XD card, and convert them to TIFFs is this feasible, or daft?
Is it better to save them on camera as TIFFs and what would the
equivalent file size be then? It doesn't matter for document copying,
what matters is to get as many reasonable frames on one card.
For other work, I have six XD 1Gb cards, so have plenty of space.
In a week away, I can easily take 1200 frames, to sift out the chaff
later ... better to grab that shot of a passing ship than to regret it
afterwards.

Best Wishes, Raymond

You won't have any problems just opening and closing jpegs to look at, etc. The problem comes when you want to change them in any way, which then requires a 'save.' It's at this stage that saving them as tifs is most useful.

With the prices of memory cards so reasonable for 1Gb and above nowadays, the main reason to choose either jpg or tif or RAW (if your camera has it) is the speed at which your camera can write the file to the card.

many modern compact digicams just have jpg available which is fine for speed and for normal sized photos is not a great problem- saving a file just once as a jpg will not be too harmful

At all opportunities I tried to isolate the image sensor to provide valid results. Once different cameras are used with different lenses results become a little skewed. It would be interesting to see how a 6Mpx camera with a high resolution lens compared with a high count Megapixel camera with a mediocre lens performs. It is something I applied some though to but decided not to include in the article preferring to keep things equal between all test examples.

Looking at my fine detail test, it would have to be a poor lens on the D2X to be anything like my 6Mpx example with a decent lens at the right aperture.

Thanks Adey, I have had this camera under a year, still
not found all the features .... at least I have everything
saved as from camera. Best Wishes

My compact is jpegs only so I always keep the original file and any alterations are done as a copy and saved in a lossless format (adobes own psd) so as to not lose anything. If I want to save space I could then convert back to jpeg which should result in no loss but still as a copy. The psd could then be deleted but I mainly just keep the psd and don't bother with the second conversion.

I will post a screen grab later.

Don

:D :D :D

Stephen,

Just a thought that high resolution sensor could be negated by poor optics. Under normal circumstances this lens would not be attatched to the X so I have no idea what I will end up with.

Don

hi Don, found the dialogue box, on "save" of image.
Sorted. Now no compression.

The camera will store in either jpeg or raw.
raw is 54 frames per 1Gb XD.

Thanks, Raymond

Interesting comments re the no of pixels etc.I took some Heron shots using the E4500,(through a scope,but makes no difference to the pixel size-I don't think-),and had them printed in an A3 size ,they looked great.

Well despite the very poor light I gave the lens test a go.
Lenses used : 55mm AIS Micro Nikkor and Stevies Sigma 28 - 200mm Aspherical IF Macro zoom. Aperture throughout f8 and the cameras were mounted on my Slik Professional tripod, and a cable release was used.

Not the best test conditions with the poor light resulting in exposure of 1/10 sec at f8 ISO 400. It is quite obvious that even at f8 the zoom is soft at the corners and overall not as sharp as the prime.

As mentioned earlier in the thread re D2X images and shrinking for posting if the image is high in detail. The full frame here is from the D100 and despite reducing to this size, it had to be compressed to 35% in 'save for web' to achieve this file size.

Don

From the brickwall test above, it seems using a high resolution lens is more important than megapixels alone.

I agree Stephen hence looking at doing the test.

When I upgraded my pc before getting the X I noticed an improvement with the D100 if I used raw and converted in NC. Next up was comparing a couple of my old manual lenses with my mid price range zoom. Again a noticeable improvement. If I had the D1X at the time, rather than the D100 which would not meter with the AIS lenses, I doubt I would have gone the route of getting the D2X. I would then have been able to afford a top notch mid-range A/F zoom like the 17-55 f2.8 and a 300 f4 to suplement my M/F lenses.

In terms of sensor and processing engine development I would expect your Hs with your glass to exceed my D100 and 55mm or 105mm in image quality. I remember when I posted my origional X or Hs thread there was a touch of ' is he mad ' Hs is only 4mp.

Don

Excellent. Quite clearly the lens is the weak link in the chain, and a quality lens beats more pixels. Not an intuitive result (or at least not for me).

I presume the crops from the D100 are 100%? Sorry if that is a dumb question.

Not a dumb question at all Leif, dumb on my part not to say :o . Yes all are 100% crops. The D2X was downsized from raw in NX before converting to Tiff. Tiffs then cropped in CS and converted to Jpegs and put onto a single frame in Paint Shop then saved at 20% compression.

Your gallery pics with the 200 micro, are a prime example of quality glass in use.

On my budget I am restricted to looking at one or two M/F classics to try and keep up with you and Foxy on the IQ front.

Don

Fascinating tests Don & Stephen and it seems to demonstrate that quality glass is the most important factor. That makes seems to me when I think about it. After all, if the lens can not resolve the detail, it doesn't matter how many pixels the sensor has, it can't recover the missing information.

This reminds me of a test I read in a hi-fi magazine back in the mid 80s (I think). They took a nominal £1000 budget and put two systems together. In the first they spent ~equal amounts on turntable (remember those), amp and speakers selecting their 'editor’s choice' products at that price point. For the second they selected a Linn LP12 turntable costing almost £800 and spent £100 each on amp and speakers again selecting editor’s choice budget products. The aim was to test the perceived wisdom of equal budget expenditure, against the philosophy that if you had a top class turntable to extract the maximum information off the vinyl record, even with a budget amp and speakers, to see which would be the more involving musical experience.

For those of you who have heard, or heard of, the legendary LP12 I won’t have to tell you the result :D

So the motto seems to be don't compromise on your lenses, if budget is an issue compromise on the camera body, and upgrade your lenses before letting the marketing hype of more megapixels persuade you into upgrading the camera. Providing they are sharp, images off a 6/8mp sensor will easily make good A3 prints.

:) :)

A good analogy Clive but I could not afford one of those either. :D Thorens TD160B MkII was more within my reach. I did lash out on the arm though on the same principle, SME Series III S. In a photographic context the equivalent of a solid tripod was through the weight of the platter and the addition of a Spectra vibration damping pad to which the LP sticks like the proverbial.

Talk of tripods or other support we have not touched on here. But a soft lens and additional blurr, however small through camera shake, and my results would have been even worse.

Don

How's about giving a clue for us ignorami (well, me anyway), who have never heard of the LP12? I presume the LP12 gave good sound even with lower grade amp and speakers?

Off topic a bit, but isn't it great that technology is constantly improving our aural experiences? DAB radio is limited to 128 kbps and often less which means that the sound quality is LOWER than analogue.

I have always maintained that it is useless having all the security features possible on a door, if the hinges are not protected, or the door is of poor construction, or that it has breakable glass in it. Same thing applies to technology. Primary first, what follows can be upgraded. Audio range is 20Hz to 20kHz, although mine was tested at 12Hz to 24kHz when I was a teenager, I doubt my range is that much now. So I don't think 128kHz should worry you unduly.
In connection with this thread ...
There are conditions in low light where I would like to use the manual settings for my DSLR, to limit the effects of noise. Can anyone offer advice as to settings to eliminate the effects of low light and the sensor - I am used to dealing with film material, which was never a problem. Should I set for 9.5Mb Fine or lower res? Should I set for a different ISO?

You bet. Linn, the makers, worked on the principle that to play a record, your base point should be to rotate it perfectly accurately at a perfectly uniform speed. The better you did that, the better the tone arm and stylus could extract the information from the record, and the more information you got off the record, the better it would sound. After all there is no point in amplifying something massively, if all you are amplifying is a mush.

A bit like enlarging a picture - creating a 20 x 16 enlargement can actually look worse if the information doesn't exist in the base image because the lens is not top quality

Sad but true - try Radio 3, Classic FM or Virgin Radio at 160 kbps and certainly I think that, providing you don't get hung up on the fact that vinyl records suffer from a few clicks & pops, musically my LP12 still sounds as good a my CD player :D

To reduce the effects of noise, my recommendation would be to first get the exposure correct. Underexposing tends to increase noise in the image shadow areas. The second thing is to use the lowest ISO you can while still having a sufficiently fast shutter speed to avoid camera shake or subject movement. As far as which file size setting to use, I would always advocate using the best available otherwise you are throwing information away. If you do end up with a noisy file, noise ninja or neat image can be very effective at reducing the effects.

I'm not sure what camera you are using - campacts with their smaller physical sensor dimensions are more susceptible to noise than DSLRs - with my Minolta A2, 200 ISO was about as far as I would go. Now with my Canon 20D 400ISO is fine and even 800 ISO is acceptable

Thanks for the info and advice, Gidders.
My camera is a Fuji Finepix S9500. Plenty of facilities
and control, but a lot more complex than using the
Nikon F2. With film, I could pretty well guess the exposure
and it would either be "spot on" or "near enough".
I have tried fog shots with the DSLR and it has been nigh
impossible to focus - I could use my eyes with the F2, but
with Digital it all depends on the sensors. Very handy though
being able to swap ISO in the middle of a session, instead
of have two cameras with different films.

Thanks. They must have uprated the BPS since I last tried a few years back ... I must try them again.

I'm clearly not an audiophile. I once bought a recording of a Shostakovich symphony (12?) which was highly rated. Indeed it was an amazing performance as the conductor and orchestra must each have snorted a few kilos of speed beforehand. But it had been recorded during the Leningrad Asthmatics Society AGM, and the constant background sounds of hordes of Russians suffering a lingering death from airway constriction destroyed the quiet passages.

I can but agree with Gidders' advice, and I will add that it is also worth using RAW. On some cameras, especially Nikon, the JPG processing reduces the image quality compared to RAW. Also, regarding noise, RAW allows you to perform noise filtering (Noise Ninja say) BEFORE sharpening. I suspect that creates better results. (This is my understanding, and not based on testing, so a little test might be worthwhile.) BTW if you do get Noise Ninja, or similar, the 16 bit version is worth having, not the 8 bit one.

Some people say you should not worry about burning out the highlights at high ISO as it is preferable to slightly overexpose to avoid noise in dark areas. But if you can increase the exposure, then why not just reduce the ISO to avoid burnt out highlights?

With my Nikon D200 ISO 800 is also acceptable, as long as I shoot RAW, and use Noise Ninja to remove the grainy effect.

In the first place I think a deal depends on the imaging programs you currently have. From an earlier post it appears you are using Paint Shop Pro 7 ( 8 bit only ) and no doubt the software that came with the camera. A look on DPR suggests that comprises FinePix Viewer & Raw File Converter LE. From that review it appears that the Raw Converter is just that, and offers no control just conversion to 8 bit Tiffs. For some odd reason it appears to convert these to 4864 x 3468 pixel, 17 megapixel TIFFs, each of which is around 50MB.

I have checked Fuji website but cannot find info on the software but I have downloaded the manual for future reference.

Given the lack of control of RAW images you might as well use Jpeg fine ( file size around 4,570KB ) and do a bit of testing at ISO 100, 200 & 400. A bit like getting to know the difference between different films, but a lot quicker and cheaper.

I have absolutely no experience of noise reduction programs so cannot really comment other than to say there was a period in the gallery of members posting and it was effective. Due to the pixel count and sensor size of the 9500 this would probably be quite a desireable program to have for ISO 200 and above.

Mainly use M/F lenses on a top flight A/F camera, almost always use a monopod or a 16lb tripod, never use NR progs, mmmmmmm, guess that may put me in the top 10 of WPF crazies. :D :D :D

Don