Thoughts of Zero One
When prowling internet forums one constantly comes across all sorts of crazy claims made about the value of cameras based on what guys (and it usually is guys) are seeing using "100% on screen” views of camera test images. I really can’t be bothered even making comment when on these forums in regards to this but I have to say it really gets under my skin.
Typical scenario, camera A with its 24 megapixel sensor is just rubbish, look at all that noise, camera B with 12 megapixels is so much better....blah blah blah. "Why do camera makers add more pixels it just makes things worse, jeeze these camera designer blokes must be idiots", and on it goes.
Now I don’t want to be rude, but this is my website so I am free to express an opinion, but these forum posters really haven’t got a clue regarding real world photographic performance.
I will admit that any image that looks sweet at 100% view on screen will likely print well so long as its done at an appropriate print size, but that does not mean that a noisy image or slightly soft image will automatically look crook when printed. In fact the noisy image may well look better than the pristine clean image in print form.
Lets just explore the problems at play here.
First of all most serious image makers don’t shoot images to simply look at them on screen, they shoot to print, I understand this is not the same for casual users who often shoot simply for Facebook or low grade online sharing. If all we wanted to do was create on-screen images then a 3 to 4 megapixel camera would be all we'd need as any resolution greater than this cannot be fitted to regular resolution computer screens without downscaling to match the lower pixel count of the monitor or device. For those lucky enough to own a 5K Mac you can see roughly a 15 meg image in its entirety but that assumes you are viewing the original image using the Macs preview option, an image on a web page is still vastly smaller and limited to the resolution of the page and browser you are using.
The Macs' 5K screen is the gold standard for hi res monitors and there is no doubt that it is a superb way to judge file quality, detail etc but still we are a long way for the point where most computer users have access to 4 and 5k screens.
For most artistic purposes the print still the final product and the final standard that we use judge image quality. Lets take a little time to look at the issues of on screen viewing as a means of judging image quality.
The resolution of computer monitors is quite variable, in the distant past monitors had as little as 640 by 480 pixels, then it was 600 by 800 and these days most are around 1920 by 1080 px, though as I mentioned some lucky folk have much higher res screens. If we assume that 1920 by 1080 px is the most common resolution still in use, we could fit approximately a full 2 megapixel image on the screen and view it at one time without down scaling, or in other words at a 100 percent view.
Screen pixel count however has little to do with the size of the screen; we can have smaller laptop screens with 1920 by 1080 pixels and larger 24-inch screen with the same resolution. Clearly then the size of the pixels will be larger on the larger screen and thus a bit coarser, this is why some high end laptop screens look so sharp, the pixels are simply much smaller and have less gaps between them (known as pixel pitch).
Monitors are very different from prints, not just in terms of the pixel versus in dots concept, but rather a monitor has a different set of attributes. Images on a monitor can display a very wide dynamic range, in other words whites can be really white and blacks very dark simply because the screen is its own light source. A print on the other hand is made on paper, it can be no darker in the blacks than the ink/dye used on the paper and no lighter than the base of the paper, additionally a print relies on the use of an external light source to provide illumination so you can actually see the reflected image, thus even the ambient viewing lighting can and usually does has great significance.
Noise in print images often has more to do with the print engine and the printer than the file itself, I have found some printers tend to exaggerate the noise present and some smooth the noise by default, sometimes differently even with the same printer dependent on the print size. I have for example printed files on Canon Pixma Printers that were for all intents clean yet still printed noisy, then made the same print again with the printers noise reduction turned on and got smoother toned prints that were sharp and noise free. The lesson, you need to experiment with your printer and realize there is a considerable amount of synergy between files, printers, drivers, papers and profiles. All is not as straight forward as that 100% on screen view may have you believe.
The colour of "on-screen" images typically looks punchier, with pure colours such as reds and yellows not being limited by issues of dyes and inks and paper bases. A monitor has little trouble reproducing the RGB colour range because it is an RGB device, but printers use CMYK or some variation of it meaning that an exact colour match for the on-screen image is very difficult but on the flip side high end prints can often show colour subtleties better than on screen images.
One of the major issues with looking at images on screen is scaling and although Adobe have addressed this somewhat with the later versions of Photoshop it is likely to remain a bugbear for some time to come. The issue is simple enough, when we have an image that has more pixels than the viewing screen it has to be scaled down by the program in use so we can fit all those pixels onto the screen real estate. The problem is scaling creates little errors and can make images look soft or jagged, over sharpened and or perhaps just plain horrible depending on the scaling percentage and the processing method employed to do the scaling. It is for this reason that the only way you can judge the clarity noise and quality of a photographic file is to view it at 100% so that scaling is removed from the equation but note that 50% and 25% views also give reasonably accurate representations and if I recall correctly, the engineers of Adobe Photoshop recommend 50% as the best view to judge the file on for printing.
Looking at the image at a 100% view is fine to judge things at a pixel level but the problem arises when the on screen image ends up very large and has little relationship to what the print size would actually be, this is exacerbated as camera resolutions rise more quickly than monitor resolutions, which is basically how it has been through the entire history of digital imaging. Take the case of the Canon 5Dr, it resolves 50 plus megapixels but as I said earlier, most people are still working with monitors of just 2 megapixels and even the very best are either 8 or 15 megapixels.
It is generally true that the higher the megapixel count of the image capture from a digital camera the higher the noise is likely to be at a pixel level, this is pretty much an accepted fact but I must say the latest crop of high megapixel compacts like the Sony RX100 series don’t necessarily support the concept as they are for the most part very clean at low ISO’s. The flip side is that higher megapixel counts generally make it possible to record vastly more detail in the first place and noise always tends to dominate if there is an absence of detail, for example in a clear blue sky. For many shots without smooth toned areas the higher noise levels may not even be evident in most photos.
Clearly a higher megapixel count also means that you can print far larger before the image starts to look mushy or lacking in detail, it may have more fine noise at those large sizes but for most purposes that is of little importance.
So here is the choice we could have a lowly 2 megapixel file that is as clean as a whistle at a 100% view but lacks fine detail and can only be enlarged to say 12cm by 18cm or we can have a 24 megapixel image that might have some noise at 100% view on screen but can be enlarged to A3 and larger without any loss of detail.
If, for example, we took the 24 megapixel image and printed it as a 12cm by 18cm it will likely look quite a bit sharper anyway and the noise is very unlikely to be an issue as compared to the actual print size the noise components are extremely small, thus you won’ t see the noise despite the fact that at a 100% view the image may look a little rough on your monitor.
If on the other hand you enlarge the small 2-megapixel file to A3 it will always be very soft and display odd artefacts and textural noise created by the interpolation process. If you then resharpen the 2 megapixel file to compensate for the loss of clarity you will most likely create odd halo effects around details as well!
Strangely enough if you really want to make your 2-megapixel image into a nice big print, and trust me it can be done, you are actually going to have to add noise to the file to even out the tonal variations and help synthesize some semblance of detail.
Finally, another aspect not properly considered is the viewing distance of the resulting print. If you take a low megapixel file and print it to a 15 by 10 cm print it will look fine and its likely your viewing distance will be quite close, probably 40 to 50cm.
If we take a higher megapixel file such as that from a 24 Megapixel camera and make an A3 print we are not expecting it to be looked at from just 50cm, the whole idea is to make it a viewable when hanging on a wall. In fact the larger we go, the greater the intended viewing distance, so any noise in the file, even if present at a 100% on screen view is very unlikely to be an issue at any appropriate distance.
Computer monitors are completely different of course because the viewing distance remains the same regardless of the megapixel count so as we enlarge the file to its native 100% max view it is always going to bias towards the low megapixel file for all the above reasons, but hopefully as you can see this has little bearing on the final intended use.
I hope this helps put the concept of prints versus on-screen view into perspective for you.