We show you how these important parameters affect the appearance of your photos.
This article is a follow-up to the one on reading lens manufacturers’ Modulation Transfer Function (MTF) graphs. We felt many readers would welcome a more detailed explanation of how these key parameters work (and interact) to influence our perception of image sharpness, whether it be in a printed photograph or an image displayed on a monitor screen.
First, some definitions.
Sharpness is quite difficult to define precisely because it’s based on subjective evaluations, which are influenced by the contrast along edges in a photo. This characteristic is known as ‘acutance’. We judge an image to be sharp when the edge contrast differences we see are relatively large.
While any photographer can recognise a sharp image, we often disagree about how sharp that image is. Different individuals perceive sharpness differently ““ and each person will make a different assessment of the sharpness of an image at different times.
Perception may be more acute first thing in the morning before the eyes become tired ““ or in the middle of the day when the brain/eye interface is at its most efficient. Assessments can also be biased by the evaluator’s pre-conceptions. We all know instances of people who have rated large and medium-format prints higher than prints from cameras with 36 x 24mm (or smaller) imaging areas, when that may not be the case.
An example of a subject and treatment (wide-angle of lens) that viewers would expect to be sharp edge-to-edge.
Sharpness also varies with different subjects ““ as do our expectations of what should appear sharp. We generally expect landscapes and shots of buildings to be sharp edge-to-edge but are much more forgiving for portraits and close-ups.
‘Soft’ (unsharp) images can be aesthetically satisfying in many cases and images in which only part of the frame is sharp can be pleasing to many viewers’ eyes.
Whereas perceived sharpness is a combination of resolution and acutance, only one of these parameters can be adjusted by photographers. Captured resolution is defined by the sensor, lens and focus in the image and can’t be changed. But acutance can be adjusted in post-capture processing.
People tend to judge images with higher acutance as being sharper, even though this is not necessarily associated with higher resolution. By increasing tonal differences along edges in an image, we can make it appear sharper, even though the actual resolution of the image may be reduced. This is how unsharp-masking works. (See right.)
Resolution defines the amount of detail in an image, regardless of how it was captured. Of the three qualities we’re looking at, resolution is the only one that is quantifiable (measurable).
The diagrams above show the differences in high and low acutance. In images with high acutance, the transitions between contrast boundaries is sharp and details have clearly defined borders. Transitions are blurred in images with low acutance, leading to a perception of lower sharpness.
Resolution is a measurement in lines per millimetre, of the smallest details a lens can resolve. The parallel bars in the top diagram are easily separated into pairs of black and white lines, providing an example of high resolution. Differences in the line pairs are blurred in the lower diagram, which can occur when a low-resolution lens is used.
Resolution can be measured in several ways. The most common method when assessing lenses is to look at how well the capture device (camera + lens) can separate closely-spaced lines. Tests of spatial resolution record the number of line pairs (each consisting of one
black line and the adjacent white space) within a millimetre width. A resolution 10 lines/mm means five black lines alternating with five white lines, or five line pairs per millimetre (lp/mm).
When looking at digital images (and camera sensors), the term resolution is often used to define the size of the pixel array. Thus, a camera that can produce images with a maximum resolution of 5184 x 3456 pixels has a ‘resolution’ of 18-megapixels. (This is derived by multiplying pixel columns by pixel rows and dividing by one million.)
For printing and monitor viewing, resolution usually refers to the number of pixels per unit length. Images that will be printed usually require a resolution of 300 pixels per inch (ppi). Computer and TV screens can typically only handle 72 to 100 lines per inch, corresponding to pixel resolutions of 72 to 100 ppi. Both will down-sample higher-resolution images.
Can an image contain too much perceived detail? Most definitely. Some subjects need to be reproduced with lots of detail; landscapes being a prime example. They are usually photographed with wide-angle lenses, and here we can apply a simple rule-of-thumb: the wider the angle of view, the more detail is required in the picture. Don’t be afraid to close the lens aperture down to the point at which diffraction begins to reduce resolution (usually somewhere between f/8 and f/16) when shooting with a wide-angle lens.
For portraits and close-ups, the situation changes and it becomes important to highlight the key feature of the subject, allowing backgrounds to blur out so they don’t engage the viewer’s eye. In portraits of living creatures (people included), the sharpest focus should be on the eye nearest the camera. The lens can be stopped down to keep the rest of the face acceptably sharp, provided the background is suitably de-focused.
In macro photography you must make a fine judgment between how much of the subject you want to render sharp and the degree of background blurring. Because depth of field is reduced as you move closer to a subject, you may need to stop down to f/8 to achieve acceptable sharpness across the subject.
Image noise may be very fine and have a very high acutance, which makes you think sharp
detail is present. However, applying noise-reduction processing will often soften image details. Careful judgment is required when applying noise-reduction processing to shots captured at high ISO settings.
Sharpness is also influenced by viewing distances. Images can look quite sharp when viewed from several metres away yet have relatively low resolution when viewed close-up. This can have a bearing on the amount of post-capture sharpening the image can tolerate before it begins to look artificially sharpened.
All digital images are slightly unsharp because of the way colour photographs are captured. The Bayer filter in front of the sensor produces red, green and blue signals that must be combined by interpolation to produce the colour image. This reduces image sharpness.
An image straight from the camera is often slightly unsharp due to colour interpolation based on the Bayer filter overlaying the sensor.
Although many cameras allow photographers to apply sharpness processing to JPEG files immediately after they are captured (and some do it by default), we counsel against it because you have very little control over the type and degree of sharpening applied. It’s better to sharpen when the image file is edited.
Minor softening from Bayer interpolation is easily corrected with unsharp masking. The process works by boosting contrast differences along edges in the image, thereby increasing acutance. The image looks sharper as a result.
Unsharp masking (and most other sharpening tools) work globally across the entire image area. For times when you don’t want everything to be sharpened, a sharpening brush allows you to apply sharpening to small areas of the image. It also works by increasing local contrast differences.
Increasing contrast can create an impression of greater sharpness.
Unsharp masking at the editing stage increases image sharpness by boosting acutance (edge sharpness).
Some other adjustments can be used to make images appear sharper. Increasing contrast will often enhance the viewer’s perception of sharpness, particularly when much of the image is unsharp. Increasing saturation may cause bright, pure colours to appear sharper than adjacent monochrome detail.
MTF and Sharpness
As we explained in the previous article, MTF (modulation transfer function) is a measurement of the relationship between resolution and contrast. The closer the 10-lines/mm curve is to 1, the higher the contrast and the better the ability of the lens to separate the line pairs. The closer the 30-lines/mm curve is to 1, the better the resolving power and sharpness of the lens.
If the MTF graph for a lens shows the 10-line/mm curve to be greater than 0.6 it’s considered a satisfactory performer. Lenses for which the 10-line/mm curve is greater than 0.8 deliver excellent image quality.
In theory, a perfect lens should be able to transmit all the light that passes through it and resolve all the black and white line pairs with neither blurring nor a loss of contrast. But, even the most expensive lenses aren’t perfect and most MTF graphs show lines that tend to curve downward as they move left to right as they track resolution from centre to corner of the frame.
MTF graphs can also be used for evaluating the potential bokeh of lenses (ie, their ability to produce smooth blurring in out-of-focus background areas). The closer together the sagittal (S) and meridional (M) lines are to each other, the smoother and more natural the background blurring (bokeh) becomes.
This is an excerpt from Photo Review Issue 56.