Noise suppression? Good cheating software!
However little light there is, you expect from a good camera that it will show what you see with your eyes. The same colors, no unnatural blur, no digital artifacts or visible noise. In practice, it’s something else. Certainly if you examine your enlargements critically.
Eyes differ from a camera sensor. And image editing in the camera or with a photo editing program has a big influence on the signal-to-noise ratio and impression of sharpness. The high image quality of modern cameras is partly thanks to sharpening and noise suppression, which starts at the pixel level on the sensor. Thanks to this advanced “cheating software,” noise is removed and some sharpness is added in every camera. Without software tricks, noise is already visible at low ISO values (below 1600 ISO), if you assess the shot at 100% on your screen. As a photographer, you do not have complete control over sharpening and noise suppression. That not only applies for jpg files, but also for RAW files. With cameras that have an anti-aliasing filter, for example, the blur that is deliberately caused by the low-pass filter by the camera manufacturer (in order to eliminate moiré) is repaired by opening a RAW file, without you being able to exercise any influence over it as a photographer.
Is this a partial enlargement of a shot made at 100 ISO without noise suppression? Or at 6400 ISO with noise suppression?
The amount of noise in a shot not only depends on the camera (brand and type, the chosen camera settings for ISO, contrast, noise suppression, sharpening and color reproduction) and the image editing (Photoshop/Lightroom) applied afterwards. The lighting circumstances under which the photo is made are also an influence. In shots made in low light or on a foggy day, you see noise much more quickly than in test shots made in the studio with a good lighting set-up.
The logical readers’ question, "Up to what ISO can I still use this camera, without having trouble from noise?" unfortunately has more than one good answer.
At first glance, you will not soon stumble across noise when reviewing your photos. Most photographers assess a small version of their shot and the enlargement at which you view the shot in part determines how much noise you get to see. The larger the photo is that you review, the greater the chance that you encounter noise. It also matters whether you print a photo or look at it on a screen. Noise will get noticed sooner on a screen than in print.
The degree to which contrast, color reproduction, sharpening and noise suppression are applied in the camera determine the image quality of the jpg shot that is stored in the camera. With some camera brands, the color reproduction, contrast and sharpening are kept as neutral as possible, while other brands choose settings for the standard image style that are less faithful to nature, but appeal to many consumers. We see those differences when testing cameras: it happens that two cameras of different brands have more or less equivalent scores if you look at unedited RAW files, but where the test results for jpg files are clearly different. That is caused by different in-camera image editing from the two camera brands. Do not go blind studying the differences in jpg image quality, since you can easily adjust those settings to your own taste. These days, you can select an image style (with names like natural, standard, portrait, colorful, etc.) on any camera that suits your taste. It is nearly always possible to further adjust these image styles on your camera, for example by adding extra sharpening, saturation or contrast. When measuring the accuracy of the color reproduction, the differences between images styles on one camera often appear to be greater than the differences between camera brands.
What is sharper: RAW or jpg?
If you open a RAW file in Photoshop or Lightroom (with the standard settings) and you compare the RAW shot with a jpg file that was stored in the camera, then you will see small differences. The performance of photo editing programs like Photoshop and Lightroom differ by camera brand. I once heard that if you apply the same degree of sharpening and noise suppression to two shots made with cameras of different brands, that the two photos will not be sharpened or relieved of noise to the same degree. In order to exclude the influence of photo editing software such as Lightroom or Photoshop, we analyze unsharpened RAW files for testing cameras and lenses, to which no suppression is applied. As you can see in the picture above, such a RAW file looks less sharp than the jpg shot that was stored in the camera at the same time. We see that in our Imatest measurement results: The MTF50 measurements of unsharpened RAW files are always lower than the MTF50 measurements of jpg files. The MTF50 of files sharpened in Lightroom or Photoshop, depending on the camera brand, is the same or higher than those of jpg files. The same applies for signal-to-noise ratio. Because there are many photographers who use the jpg files directly out of the camera, we test jpg files. We therefore use the standard settings of the manufacturer (only for testing the color reproduction we sometimes choose other settings).
In the picture above (image excerpts from a RAW and a jpg shot that are stored simultaneously in the camera), you see that the contrast of a jpg file is visibly greater than of an unsharpened RAW file. At 100 ISO, you still see little difference in signal-to-noise ratio. Not only the contrast, but also the resolution (“micro contrast”) of the 100 ISO jpg file is clearly greater than that of the unsharpened RAW file, as you can see in the example below:
Even so, you cannot say that RAW files are always less sharp or always sharper than jpg files. It depends on the camera settings and the RAW software used for editing the RAW shot. Because a camera has much less calculating power than a computer and photo editing programs therefore use much more advanced (time-consuming) algorithms than cameras do, you can always get better results with careful RAW editing than from the jpg files that are stored in the camera. Yet modern cameras are so advanced that you have to have a good amount of experience with editing before you are better than the camera. And even then, the differences are sometimes so small that many photographers do not find it to be worth the trouble and happily use the jpg files.
The higher the ISO, the more noise?
The less light there is, the higher the ISO value you have to set for the camera in order to prevent underexposure or motion-blurred shots. But the higher you make the ISO value, the more noise becomes visible. At least if you do not apply any—or precisely the same amount of—noise suppression. At 6400 ISO in the partial enlargement of the unedited RAW file shown above, it can be clearly seen that the signal-to-noise ratio at 6400 ISO is already greatly decreased relative to 100 ISO. At the lowest ISO settings, the differences are still small, but at the highest ISO values, the noise is more dominantly present. Above 6400 ISO, it goes quickly. At 25,600 ISO, the unedited RAW file is no longer usable. High time to call in the help of our favorite cheating software: sharpen and suppress noise.
In-camera noise suppression (jpg) or Lightroom/Photoshop noise suppression (RAW)
Above you see two 100% partial enlargements of test shots. In the jpg shot, you do not see color noise anymore, but a grainy patter (“noise”). Thanks to noise suppression and sharpening, it even appears as though the 25,600 ISO does not produce a usable shot, if you only look at the global contrast. But if you look at fine details, the picture is less rose-colored:
Detailing of a shot suffers under noise suppression and sharpening. From artifacts in a model’s hair, or even hair in which every detail has been lost, you can clearly recognize the effect of excessive noise suppression or sharpening. The subject of a shot has a big influence on the image quality. On an even surface, the noise becomes noticeable sooner, but if it is really even, then noise suppression software is able to mask that for a long time. At high ISO values, the noise suppression software is more successful at preserving the color of large areas (like the bird above) than for details in the shot below. If a shot includes flowing lines or geometric patterns, then noise suppression and sharpening are able to preserve those patterns well at high ISO values. Above you see two partial enlargements, shown at 100%. The bird in the edited 25,600 ISO RAW shot (standard settings of Lightroom: 25 sharpening and 25 Chroma Noise) look more natural, but you do see a disruptive digital noise wash across the shot. Nearly everyone finds digital, pixelated noise less attractive than organic, grainy noise. The standard settings in Lightroom suppress only color noise. With a small amount of noise suppression, this RAW shot can be significantly improved. The jpg shot (directly out of the camera) is edited by the camera in such a way that there is much less of the bird left.
With irregular patterns with fine details, such as a colored marble floor, it is no longer possible for noise suppression to cheat. To illustrate that, we took a shot of a test card with an irregular pattern:
When is the limit reached?
The 25,600 ISO RAW partial enlargement (left) above shows where the limits for cheating software are reached. If you compare the 25,600 shot with the 100 ISO shots at the top of the page, then it is clear that a great many details in the original disappear in the noise. You can no longer magically repair this by means of noise suppression or sharpening.
But that is not all. Noise suppression software sometimes mistakes the color of fine details for color noise (Chroma noise). The shot on the right shows that the color in the details in the jpg (not shown here, but also in the RAW file that is developed with standard settings in Photoshop or Lightroom) have completely disappeared.
Cheating becomes magic with PRIME from DxO Optics
The comparison above of an unedited 100 ISO shot with an edited 25,600 ISO shot shows what the best noise suppression software of today is capable of. These are partial enlargements that are shown at 100%, and you will usually review a smaller illustration, and then the remaining noise that you still see in the shot on the right is no longer noticeable. But if the global contrast of a shot is good, that does not mean that the 25,600 ISO shot looks just as good as the 100 ISO shot. With the 25,600 ISO shot, noise suppression software could preserve the green color of the bird, but if the pattern is less predictable, then DxO Optics Prime noise suppression (only applicable to RAW files) is the package that offers the best results for what I have tested so far.
The best method for making perfect shots is to choose the best equipment possible and to use the lowest possible ISO values and a tripod. That is different in practice for many photographers. We usually photograph by hand and then need a higher ISO value in order to prevent motion-blurred shots if it isn’t bright, sunny weather. Fortunately, there are ever-more affordable, bright lenses, cameras keep getting better, and image stabilization ensures sharp pictures if you choose a longer shutter time. All those developments make it possible to choose a low ISO value even without a tripod, without having to be afraid of motion-blurred shots.
If you have taken a picture of which you are very proud, but where there appears to be too much noise with enlargement, then you can still book 1 to 2 stops’ profit with advanced noise suppression software like DxO Optics, Noise Ninja, or Neat Image, without losing too many details or disturbing artifacts becoming visible. That actually only applies if you still have the RAW file. A jpg file has been edited so that it really isn’t worth the trouble of editing it any further.
Is this a partial enlargement of a shot made at 100 ISO without noise suppression? Or at 6400 ISO with noise suppression?