Why Your 21MP File “Looks Softer” Than Your 12MP File at 100%

Posted on 01 May 2011

Another photography forum meme is that “my old camera with fewer photosites produces sharper images than your newer camera with more photosites. In fact, I have carefully inspected 100% crops from both and the evidence is clear.”

Not so fast.

Let’s imagine that the comparison is being made between the full-frame 12MP Canon 5D and the full-frame 21MP Canon 5D2, both of which I own and use. If you put the same lens on both cameras, set the lens to the same aperture, and point both cameras at the same subject, the lens will project exactly the same image onto the sensors of both. Let’s say that you do this under controlled conditions, and you decide to compare the two captures to see if the 21MP camera is really sharper than the 12MP camera.

You want to compare closely, so you display both images as “100% magnification crops” – portions of the image that show each individual pixel from the original photograph as an individual pixel on your screen. You display the two images side by side and, squinting closely and looking back and forth between the two, you notice that the 21MP original is certainly no sharper than the 12MP original and that the 21MP image actually looks a bit less sharp! You decide that a) higher MP cameras are less sharp than lower MP cameras (you have the evidence!), and/or b) the camera companies are pulling a fast one on us.

When gazing too long and too intently at a computer screen, it is all too easy to forget that the real world is not always represented accurately on the screen. In this case, the error is a result of viewing on a computer screen rather than making more realistic comparisons, for example between two prints of equal size. With the screen image comparison, you might overlook a fact that explains why the sharper (or at least equally sharp) camera appears to be less sharp. A look at the above  illustration will help.

The image includes two copies of a full-frame photograph. Think of the one on the left as representing the photograph made on a 12 MP camera (like the 5D) and the one on the right as representing a 21MP camera (like the 5D2). The original 5D image would be 2912 pixels wide, while the original 5D2 image would be 3744 pixels wide. The full-color area of each image represents the part of the original image that would fill the screen of a 1280 x 1024 monitor when the originals are viewed at 100% magnification.

The important point that is illustrated here is that the 1280 x 1024 “slice” of the 21MP image shows a considerably smaller portion of the overall image, and in order to fill the same size screen it will have to be magnified more than the image from the 12MP camera. If the two images are equally sharp to begin with, the one that has to be magnified more to fill the screen will lose more of its original resolution because you are looking more closely at a smaller portion of the image.

In the end, if you were to make two prints of the same dimension from the two original images, the higher MP original would look at least as sharp as the lower MP original, and if you use good lenses and good technique (and print large enough that it makes a difference) the higher MP version has the potential to resolve more detail.

(See related post: “Myth: Diffraction and Motion Blur Worsen With More Megapixels”)

G Dan Mitchell Photography | Flickr | Twitter (follow me) | Facebook (“Like” my page) | LinkedIn | Email
Text, photographs, and other media are © Copyright G Dan Mitchell (or others when indicated) and are not in the public domain and may not be used on websites, blogs, or in other media without advance permission from G Dan Mitchell.


7 comments to Why Your 21MP File “Looks Softer” Than Your 12MP File at 100%

  • [...] Why Your 21MP File “Looks Softer” Than Your 12MP File at 100% [...]

  • Someone tweeted in response to this article and suggested that differences in diffraction blur and/or greater sensitivity to camera motion might explain why images from higher photosite density camera really are less sharp.

    I’m afraid that neither is the case.

    Diffraction blur is an optical quality, not a sensor quality. As you stop down a lens, at a certain point the maximum resolution begins to diminish due to diffraction blur. A “point of light” expands to cover a larger area of the sensor. On a cropped sensor camera you might become concerned about this if you stop down much past f/8 – at least if you are very critical about resolution and perhaps intend to make a very large print. On a full frame camera you might become concerned roughly two stops later, perhaps beyond about f/16.

    The idea that diffraction blur increases as you increase photosite density (e.g.- use “more megapixels”) is rooted in the very same misunderstanding that I described in the original post – failing to take into consideration differences in the way that images are seen at 100% magnification on the screen versus how they are seen in real final images such as prints or on-screen jpgs. Imagine some very gross diffraction in which (to use loose terminology) the “blur” from diffraction is 1% of the width of the frame. (This would be absolutely horrible diffraction blur, and it is far beyond what you’ll see in the real world – but 1% is a nice convenient value for this explanation.) Since the lens produces the blur, not the sensor, this “1% blur width” will be the same whether the image is directed at a piece of film, a 8 MP sensor, or a 21MP sensor. In fact, for the thought experiment, imagine that you make photographs with all three media. Now make three prints at whatever size you prefer – let’s say 16″ x 24″ for the sake of having a real size in mind. The “1% blur width” will be 1% of 24″ in all three of the prints. In other words, there is no difference in the amount of diffraction among the prints due to different recording media or different photosite densities.

    The situation with motion blur is essentially the same. The crucial issue is over what portion of the image the blur takes place. If it is, say 1/10,000 of a frame width the blur will be 1/10,000 of the print width no matter what number of photosites you use – ignoring for a moment the fact that no current full frame DSLR can resolve 1/10,000 of the width of the frame. But let’s say the motion blur is grosser – perhaps 1/100 of the frame width. It will be 1/100 of the picture width in all three cases, independent of the film/sensor characteristics.

    The place people seem to get confused is, yet again, when they look at 100% crops on their monitors and, again, forget that they are looking more closely at a smaller portion of the image when they view the image from the high photosite density camera. The blur would, indeed, cover a very slightly larger percentage of their monitor width but this is exactly counteracted by the smaller portion of the image they are viewing.

    But, one of them says, “the blur” (whether motion or diffraction caused) “covers more pixel widths!” Right. It does. But that doesn’t change the fact that the size of the blur relative to the size of the print or other final display is no larger at all.

    To those who imagine or suggest that higher photosite density creates a “diffraction-limit problem” or “need for greater camera stability to control blur” (some have gone so far as to suggest the use of higher shutter speeds on higher photosite density cameras!), there is a very good chance that you are accidentally thinking of something as a problem when in reality it is either neutral or an advantage. You will get exactly the same amount of difffraction blur or motion blur in a print of a given size with any of the available photosite densities. In the best case you might get slightly better resolution by opening up a stop or so if you are using a very good lens and you are extremely careful about focus and camera stability and if you make a really big print.

    In terms of diffraction or motion blur, when it comes to the photographs you produce with your camera there is no bad news and perhaps at least a bit of potential good news when you move to a camera with higher photosite density.

    (Disclaimer: I am not proposing that everyone need higher photosite density, nor that everyone will seen improvements in their photographs if they move to a higher MP camera. There are many other factors here than might negate advantages, but few if any that are real disadvantages.)

  • [...] Diffraction and Motion Blur Worsen With More Megapixels A recent post (“Why Your 21MP File “Looks Softer” Than Your File at 100%“) explained why pixel-peeping* photographers might imagine that cameras with greater [...]

  • I think you are correct, but likely the misunderstanding comes from the increased noise of some of the higher-megapixel cameras.

    That is why Canon went from nearly 15mp on the G10 to 10MP on the G11/G12. I am sure that in due time, sensor technology will improve and the count will go up again.

    At least with more pegapixels you have more chances to see the imperfections of your lens ;-)

    Cheers, Harry

    • Thanks for visiting and posting!

      Your point about the relationship between noise and higher pixel density in an interesting one. Theory (and many forum posters!) have held for a decade or more that increased photosite density would result in increasingly noisy photographs. However, if these predictions were accurate, as photosite density has steadily increased from 4 to 6 to 8 to 10 to 12 to 20+MP we would have seen very significant degradation in noise performance. But it has not happened. There are a range of reasons for this. Software has evolved to deal more effectively with noise in-camera, as has post-processing software for those relatively rare situations in which it must be dealt with in post. The space between photosites has decreased, allowing increases in the number of photosites without an equivalent shrinkage of the individual sites. There have been increased efficiencies in the photo sensors themselves.

      In any case, noise and sharpness are quite different things. While noise is not an unimportant consideration when it comes to image quality, it doesn’t really have any effect on the issue I’m addressing in this post, namely the failure of some viewers to keep in mind the effect of photosite dimensions on what portion of the image they are looking at when they view at 100%.

      The issues with the very small cameras like the G series from Canon are two-fold I think. First, because they are really pushing the boundaries of minimizing the size of individual photosites, there are issues with maintaining a high enough signal to noise ratio. It works quite well at the lowest ISOs, but the normal noise-enhancing effect of higher ISOs becomes acute much sooner with the small sensors. But there is another factor as well. While you can put a 15MP sensor in such a small camera, in order to use that sensor resolution in any meaningful way your lens is going to have to be quite extraordinary… and shooting technique will need to be nearly perfect. Those two issues do not fit well with the market into which $400-$500 small “cams” are sold. No manufacturer can produce a lens that competes with or is better than the best Canon L (or equivalent) lenses and put such a lens in a relatively inexpensive body. And most people buying these cameras (including me! I have the S95) choose them for shooting that is casual.

      Of course, you said as much when you astutely observed, “At least with more pegapixels you have more chances to see the imperfections of your lens ;-)”

      Lots to think about!

      Dan

  • JohnNo Gravatar says:

    So can it be argued that if we can determine the equivalent zoom percentage to show the same screen portion as the 100% view of the lower res image, the higher-megapixel image should look as sharp or sharper? And the upshot is that since you are looking at a smaller section of the image when reviewing at 100% you are really seeing the limits of the resolution of the lens you are using, not the effects of higher pixel density?

    • Yes, that would be the case – “as sharp” in terms of lens performance, and possibly “sharper” if the lens performance exceeds sensor limitations.

      A better way to think about it may be to consider your final output. If you print at a given size – lets’s think big and go with 24″ x 36″ – when all else is equal the print from the higher resolution sensor will never be worse than that from the lower resolution sensor and it may be better…. even in a case when a 100% crop inspection might make it appear otherwise.

      Dan

  • Leave a comment



    Optimization WordPress Plugins & Solutions by W3 EDGE