As camera manufacturers are continuing the megapixel race, with Sony releasing a bunch of 24 MP APS-C (1.5 crop-factor) cameras like Sony A77, A65 and NEX-7, and Nikon releasing a high resolution 36 MP Nikon D800, many of us photographers question the need for such a high resolution sensor. Some of us are happy while others are angry about these latest trends. Just when we thought companies like Nikon abandoned the megapixel race, instead of seeing other companies do the same, we now see Nikon back in the game with a new breed of product with a boatload of pixels. Why did Nikon all of a sudden decide to flip the game? Why does everyone seem to be going for more pixels rather than better low-light / high ISO performance? Does a high resolution sensor make sense? What are the true benefits of a high resolution sensor? In this article, I will provide my thoughts on what I think has happened with Nikon’s camera strategy, along with a few points on benefits of a high resolution sensor.
Pixel Size, Pixel Density, Sensor Size and Image Processing Pipeline
OK, this topic is rather complex if you do not know anything about pixels and sensors. Before you read any further, I highly recommend to read my “FX vs DX” article, where I specifically talk about pixel and sensor sizes and their impact on image quality.
As you may already know, pixel size, pixel density and sensor size all contribute to how well a camera deals with low-light situations (high ISO performance) and how well it sees range of light (dynamic range). Pixel size is a very important attribute of a sensor’s overall performance – typically the larger the pixel, the better the overall performance. Pixel density is closely related to the pixel size – larger pixels equal lower pixel density, smaller pixels equal higher pixel density. That’s because pixel density is measured by the number of pixels per inch. There is a fourth, very important attribute that very few people mention when talking about pixels and sensors that also plays a huge role; it is the software algorithm run by the image processor that analyses the data from the sensor and runs a series of image processing steps to reduce various artifacts, reduce noise, apply sharpening and more. This is commonly called the “image processing pipeline”. All four of these factors significantly impact the overall image quality and are closely related to each other. A good camera should have a good balance of pixel size and pixel density, sensor size and image processing pipeline.
Let me give a few examples to clarify this a little more. If you have two identically sized sensors – one with small pixels (hence higher pixel density), and one with large pixels (lower pixel density), everything else being the same, the former should generally produce lower quality images than the latter, especially when it comes to noise. The Nikon D3s, having much larger pixel size performs much better at high ISOs than the Nikon D3x (when viewed at 100%), which has more pixels / resolution and smaller pixel size. Makes sense, that’s why Nikon makes two different cameras for different needs.
Now let’s take another example. If you take two cameras with different sized sensors, “A” being the one with a larger sensor and “B” being the one with a smaller sensor, which one would perform better? It would depend on pixel size and density and the image processing pipeline – the other important variables I talked about above. If the image processing pipeline is exactly the same and the pixel size on camera “B” is the same as in camera “A” (hence “B” has less total resolution), then we should see very similar pixel-level performance. Now what if camera “B” has the same resolution as camera “A”, but has a much better image processing pipeline? Pixel size on camera “B” is smaller, which technically should make camera “B” produce more noise, but its image processing pipeline is superior and hence it compensates for the difference. When comparing images from both cameras, despite variances in sensor sizes, you might see very similar noise performance (I am obviously excluding depth of field and other differences for simplicity purposes). I explained this in more detail in my Nikon 1 V1 Review. While having a much smaller sensor than the competition, the Nikon 1 V1 shows impressive high ISO performance due to a much better image processing pipeline. When people first saw that the Nikon 1 V1 high ISO images look clean, many claimed that Nikon was “cheating” by adding noise reduction at high ISOs even on RAW files. What they don’t realize is that Nikon has been doing it for a while now and it is by far not the only manufacturer that does it. Everybody is doing it nowadays; otherwise images would look too darn noisy! There is absolutely nothing wrong with this sort of noise reduction, as long as the manufacturer knows how to properly apply noise reduction without losing too much detail.
Lastly, let’s take two different cameras with identical sensors with the same pixel size and density. One might perform better than the other in terms of noise. How? Again, better in-camera image processing. Sony manufactures most of Nikon’s sensors and uses those same sensors in their Sony Alpha DSLRs. And yet due to Nikon’s better image processing pipeline, Nikon cameras show better overall image quality, specifically at high ISOs. Same sensors, different output.
There are other important variables such as overall quality of sensor, bayer and anti-aliasing filters that also contribute to overall image quality, but I am not adding them to the mix for simplicity purposes.
Nikon’s Change in Strategy
So why did Nikon all of a sudden decide to reverse its game and go with a high resolution sensor on a lower-end full-frame body like the Nikon D800? Because it makes sense for Nikon. Canon realized this a while ago, which is why it introduced the Canon 5D Mark II with a 21 MP sensor. Nikon started out with its flagship Nikon D3 line, then came up with a lower-end D700 body that used the same sensor, same AF and other specifications, including the image processing pipeline. As expected, the lower-end Nikon D700 started to heavily cannibalize the D3 sales. Demand for the D700 skyrocketed, while D3 was not selling so well anymore. Then Nikon released the D3x as its flagship “high resolution” camera. With the pricing strategy Nikon chose, it killed the potential D3x sales and made it out of reach for most people out there. By then, Nikon D700 was selling strong and both D3 and D3x were suffering badly. Then came the Nikon D3s, which offered significantly better low-light performance. The flagship product was back in the spotlight and sales figures started to look better – those who needed the best camera would get the D3s, while everybody else that had budget constraints had to live with the D700. D3x continued to suffer, despite the drop in price. Meanwhile, Canon was doing really well with its two cameras – the Canon 5D Mark II sold like crazy, while pros that needed better low-light capabilities got the 1D Mark IV (if only it was not for the plagued AF issues on the 1D series, the camera would have sold even better). What happened with Canon 1Ds sales? That’s right, just like the D3x they also plummeted. Canon 5D Mark II has been the best selling Canon camera among most photo enthusiasts and pros. Differences? Canon 5D Mark II appeals to all kinds of photographers from wedding/events to landscape and fashion photographers, while the D700 mostly appeals to wedding/events and sports/wildlife with the MB-D10 battery pack. The last part is where Nikon made a mistake. With the MB-D10 and the right batteries, the Nikon D700 can be almost as fast as the original D3, sharing mostly identical features and costs much less. Nikon D3s was late in the game – D700 sales were still very strong even after D3s came out. We all anxiously waited for the Nikon D700s with the same D3s sensor, but it never came out. Then we thought we would get a D700x, which also never materialized. If Nikon released a D700s, it would have killed D3s sales. If it released a D700x, it would have buried the D3x forever.
Now we are about to see a significant change in strategy, with a new breed of a lower-end D800 pro line with a high resolution sensor. Do the same thing Canon does – high resolution, low FPS, perhaps less features here and there to differentiate it from the D4 line, so that it does not cannibalize the D4 sales. Sports and wildlife photographers spend tens of thousands of dollars on expensive 600mm lenses, so they clearly can afford to buy the D4. Everyone else that cannot, has to live with a camera that would appeal to a large group of people – from landscape, architecture and studio photographers to event photographers that do not seem to mind a high resolution camera. Canon released the 7D to compete with the D300s and recently introduced the 1DX to compete with the D4, why not bite them back with something that can challenge the 5D Mark II?
But revenge and larger market capture are not the only reasons why Nikon decided to go with a 36 MP sensor on the D800, in my opinion. There are two more key factors here – high resolution sensors are cheaper to make in the long run for Nikon than low-light sensors. Sounds wrong, but Nikon spends a lot of R&D money on its noise reduction algorithms. And after spending all that time and money, it is painful to see something like the D700 cannibalize its flagship line sales. Do you know that both Nikon D3 and D3s have almost identical sensors? The difference between Nikon D3 and D3s is mostly software – the same image processing pipeline changes I have been talking about. That’s why you do not see any improvements at low ISOs – the Nikon D3s only looks better above ISO 800. With Sony making high resolution sensors for Nikon, it is better to just go with the flow than continue the same trend. Put the primary focus on the flagship line, make it super attractive for all that need it and can afford it and introduce a lower-end pro line for everyone else that wants a high-resolution but slow full-frame camera. Increase the price on the latter so that it does not eat up the flagship sales and the problem is solved. Nikon knows that Canon abandoned its 1Ds line, so why bother with two flagship products? Most likely we won’t be seeing a D4x in the future.
This is my analysis of the current Nikon situation. I might be wrong, so we will see – time will show.
The Benefit of a High Resolution Sensor
By now you have probably read about the “megapixel myth” and have probably heard this phrase a number of times: “camera resolution does not matter”. It sure does. Now before rotten tomatoes come flying my way, let me first finish the sentence: depending on what you do with your photos. If you only publish your photos for the web, or print on your small printer at home or provide pictures to your wedding/event clients, then you would rarely need more than 10-12 megapixels. But if you are a landscape or a fashion photographer that wants to sell large prints then you need a high resolution camera. Even many wildlife photographers choose to shoot with a DX camera for the “reach”. If I could get my hands on a 36 MP sensor and have the ability to crop my frame to what I can get with a DX camera today for the same “reach”, I would be a happy camper. There is a reason why there is demand for high-resolution cameras like the Canon 5D Mark II. If digital Medium Format cameras were affordable, those 40-50 MP cameras would be in huge demand. Yes, most of us can easily live with a low-resolution camera. In fact, considering what most of us end up doing with our pictures, even the compact mirror-less cameras would suffice for 90% of our needs. However, there are many photographers out there that would hugely benefit from a higher resolution camera.
The megapixel myth is true, but it works both ways. The need for a high-resolution sensor is as much of a myth as the need for a low-light sensor. I own the Nikon D3s, which has been the king of low-light photography until D4 came out. How many really high ISO images above ISO 1600 do you think I have delivered to my clients, compared to low ISO images? Not that many. Why? Because to get the highest quality images, I avoid shooting at very high ISOs when possible. Except for some extreme situations like shooting wildlife at dusk or dawn and maybe shooting in dim churches, you will find yourself rarely using extremely high ISOs. If you don’t believe me, open up your Lightroom and do a quick count of images below ISO 800 and above ISO 800 for the last year. You will be surprised to see the numbers (unless you have no idea how to shoot your camera and have your D3s permanently set to ISO 3200 :))
My point is this – a good general-purpose camera should have a good balance of sensor resolution and low-light capabilities. So, extremely high resolution is bad, because image processing algorithms won’t be able to cope with that much noise today. And at the same time, you are not getting much out of a low-light sensor if it has too little resolution.
Finally, what is the benefit of a high resolution sensor? The true benefit of a high resolution sensor when compared to a low-resolution / low-light sensor, is that you have the option to down-size/down-sample your images in post-processing. Why would you want to do that? To reduce the amount of noise, of course. With a high resolution sensor, you have the option to shoot really big images and you can down-sample their size in low-light situations to reduce the amount of noise. With a low-resolution low-light camera, you can shoot good quality images right out of the camera, but you can never do a great job of increasing its resolution (as that article demonstrates, the best you can do is make minor improvements). You would be surprised how little of a difference you will see when taking an image from the D3 at ISO 3200 and comparing it to an image from the D3x at ISO 3200, down-sampled to 12 MP. Now I am not here to say that D3x and D3 have the same high ISO noise, because it depends on how you look at it. At 100% pixel size view, the Nikon D3 is obviously going to look better. But when both are viewed at 12 MP, meaning the D3x image down-sampled to 12 MP to match the D3 image, then the images will look similar in terms of noise. Don’t believe me? Head on to DxOMark, put D3 and D3x side by side, then look at the SNR chart under “Measurements” in “Print” size:
Those noise levels look about the same to me. If Nikon D3x came out after D3s, we would have probably seen similar results in the above chart when comparing the two.
Is 36 MP an overkill on a full-frame sensor? Given what Nikon has been doing with its noise reduction, then no, it is not an overkill. If my projections are right, we should be seeing at least a 1 full stop improvement over the D700 when the D800 image is down-sampled between 12-16 MP. What this means is that at ISO 3200, the Nikon D800 should look about the same or better than Nikon D700 at ISO 1600, when the D800 image is down-sampled to 12-16 MP. And you should end up with a sharper D800 image on top of that (due to resizing). Nikon’s built-in noise reduction, along with Photoshop/Lightroom software image reduction algorithms should make this happen.
Don’t be scared of 36 megapixels. Remember, the pixel size on the D800 will be the same as on the current D7000. If you find your lenses to be working well with the D7000, they will work equally well on the upcoming Nikon D800, with the exception of corners – that’s where you might see differences. That’s because D7000 hides the corners of full-frame lenses due to the smaller sensor, while the D800 will expose them in full. But you can work around those problems in the field. Crop the corners a little more if your lens has bad corner performance – you will have plenty of pixels to work with.
A high resolution sensor would obviously have its own disadvantages as well. More resolution equals bigger files, hence slower FPS (I know some might say slower post-processing as well, but it is debatable, because computers are very fast nowadays and should be able to easily cope with higher processor and storage requirements). On top of that, even a large memory buffer would clog up pretty quickly, so shooting continuously at 4 FPS for more than a few seconds is out of the question. But that’s what the D4 is for. Need to spray and pray? Get a D4. And if buffer or image size are such a problem, shoot away in DX mode. You will have that option on the D800.
What are you thoughts on all this?
You wrote about downsizing of images to reduce noise. Well, I am not an professional photographer, but sometimes when I downsize an image, some pasterns, e.g. parallel lines that are not completely horizontal or vertical, become awful (particularly when the ratio is 66.6% or 33.3%).
So I wonder if you could analyse this technically and practically.
I appreciate the fact that you are presenting information for all levels of photographer and not ‘talking down’ to the amateurs. Really useful information. I am considering trading the D700 up to the D800 as a back-up camera. Currently shooting the D4 and it is a tank! I shoot mostly sports – rodeo, motocross – and throughout the dust, mud, pouring rain and lousy indoor arena lighting it has NEVER failed to perform.
Based on this logic, would a high density sensor in ASPC model would have same quality image than the low density full frame sensor? Thanks
“What are you (sic) thoughts on this?”
a) I don’t care about Nikon’s marketing strategy.
b) If your technique is not already impeccable, then a Tripod will do more for your sharpness and REAL resolution than any number of extra megapixels. Hand-held, in bright light with very short exposures and a good IS/VR/OS lens you might get away with it. Following the old rule of (exposure inseconds) = 1/(focal length in mm) … not a chance.
With a full frame sensor the linear pixel density of the likes of the Nikon D800 and the Canon 5DS is in the region of 200 sensor elements per milimetre, or even more. It means that if the sensor moves so much as 1/50mm during the exposure you’ll have effectively reduced the TRUE resolution of your 36Mp or 50Mp sensor to less than you can capture with a 6Mp sensor on a sturdy tripod. 1/50 mm is a hardly perceptible movement.
In other words a photographer that takes the trouble to get the most out of their equipment will still capture more real detail, and capture it more sharply, with something like an EOS 10D (6Mp) than most amateurs (and more than a few professionals) can capture with the latest and greatest equipment.
Thanks for the insight and education.
Sincerely appreciate your opinion amongst the spectrum..
Wonderful article thanks for bringing up downsizing. I will look into downsizing my Sigma Merrill images. More time in front of the computer!!
Very useful article, thanks.
Do you think the D7100 24MP (roughly equivalent to 50+ MP full frame) ‘outresolves’ the 300mm f4D ED IF?
Very useful, thanks.
I have a question that is related to down sampling. When you take a picture and down size it – the software does the calculations and either throws some out or averages them in or a combination of each. What happens when you take a 16 megapixel camera and set the camera to 5 megapixels. Does the camera still take a 16 megapixel picture and use software to down sample it or does it still use the entire sensor but make the pixel size larger ( I heard that pixels have no physical size, that it is a unit of measure derived from division ).
Sorry for posting to this old threath, but I discovered it only today and I found it very interesting.
Nobody answered this interesting question posed by Jeremy: “However, does setting your jpg file size from L to M or even S produced a noise reduction effect?”.
I have the same issue and would like to get a response by you the experts.
Many thanks in advance.
Alessio, an old thread indeed! Yes, reducing JPEG file size to M or S does reduce in lower noise due to in-camera down-sampling.