Sharpness is an essential quality of images nowadays. If it’s not sharp…. well, it may have been acceptable 50 years ago . Today it’s just not on. Modern cameras and lenses can achieve incredible sharpness and software programs can enhance this further. The results can cut your eyelids. Sharpness has become the new norm.
Principles: A theory of sharpness
- Sharpness is an optical property. It happens when you focus light through a lens. In fact there is only one distance from the lens where the image will be truly sharp. This corresponds to the point of focus for the given lens. (not the focal length). The higher quality lenses are often sharper than cheaper lenses as they compensate for things like chromatic aberration. Autofocus does not improve sharpness, but may have improved the ability of photographers to focus on a given point.
- The region around the focus point that looks sharp to the eye is known as the depth of field. If you study an image with a magnifying glass you will discover that the depth of field loses its sharpness when magnified. Lenses that magnify (telephoto and Macro lenses) have very narrow depth of field. Conversely wide angle lenses have a greater depth of field. (Also worth mentioning that the depth of field improves with narrower apertures.)
- The exception to the above principle is the Frazier lens system. This was developed by an Australian for macro cinematography with an almost limitless depth of field. These lenses are very useful in trick photography.
- Acutance is the term given to the form of sharpening that enhances edges. The technique involves negative feedback to surrounding pixels. Homogeneous areas will remain uniform while edges will be enhanced – the dark side becoming darker and the light side becoming lighter. This can be done mathematically by a computer. Prior to computers it was achieved in the dark room by overlaying a mildly blurred positive transparency over the original negative. Interestingly this process has also been found in the retina at the back of the eye. It was discovered that the retina has a layer of horizontal fibres that provide negative feedback to surrounding rods or cone cells, creating it’s own edge enhancement.
- Deconvovulation is the term given to a computer algorithm that attempts to undo blurring and reconstruct the sharp image from a defocussed lens. It was used to improve images from the Hubble space telescope when it was discovered that its mirror was the wrong shape. (too shallow) Deconvovulation requires knowledge of how the image became blurred. It can attempt to undo motion blur, lens blur or gaussian blur. Unfortunately it can produce a lot of unwanted artefact.
- Lens Blur occurs when the lens is not properly focused on the image. It will turn points of light into circles.
- Motion Blur occurs when the camera or the object move while the image is being made. It will turn points of light into lines.
- Sharpening may produce unwanted side effects in the image. These include halos around borders, increase in noise and sharpening artifact. Sharpening artifact is the rendering of textures as a high contrast snow, like the picture on an old TV.
- Terry Branford informs me that digital compression inevitably results in some fine blurring in the order of 1-2 pixels. He recommends a fine sharpen of all digital images in this magnitude to get more realistic results.
Strategies that will reduce Sharpening artefact:
- Unsharp Mask
The unsharp mask filter will enhance only some edges. It chooses those with the greatest difference between the dark and the light side. It will leave edges with small differences unaffected. The cutoff point or threshold can be chosen by the user. The effect will result in less noise generation in the smooth background areas. Unfortunately this strategy does not prevent halos or the generation of artifact in areas of strong texture.
- Selecting what to sharpen
Most people prefer images that isolate the subject from the background. It is pleasing to have the subject sharp and the background blurred. In reality only some of the subject will be sharp and some of the background may be discernible. Rather than sharpen every pixel in the image, it may be wise to select out those areas that need sharpening. In general you should sharpen the slightly blurred areas in the subject – and leave the background unsharpened. This can be done by using the selection tools in photoshop. My preference is to work in layers and using masks to select which areas to sharpen and which to leave unsharpened.
- Remove noise with a noise filter
This is my last – not my first option. Noise filters can result in loss of detail. The best noise filters aren’t in Photoshop. I personally use the NIK plugin filter, Dfine 2. There are many other filters available. My general advice is to remove digital noise before you sharpen. Noise is a lot harder to remove after it has been sharpened. However if the sharpening produces a lot of additional noise, you may wish to run it again after sharpening. I must warn you that you may lose some detail doing this. I tend to use the selection tools to limit the areas that need most noise reduction, so as not to lose detail in the subject.
How to sharpen an image
- Choose the areas that need sharpening
Don’t sharpen bokeh. Don’t over sharpen areas that are already sharp.
- Analyse the cause of the blurring.
For instance is it motion blur or lens blur? What is the diameter of the blurring? What is the direction of movement? Use the ruler tool to measure the length and direction of blur lines or the diameter of blur circles.
- Use layers if possible
This allows you to modify the areas to be sharpened (by using masks) and to reduce the severity of sharpening (by blending it back with the original layer).
- Select a sharpening strategy
Unsharp mask is sufficient for most purposes. Some people prefer High pass filter (look in the “Other” menu) as it can be easily applied to layers. Smart sharpen may be useful for motion blur or higher level lens blur.
- The settings in the sharpening filter.
Most importantly you must select an appropriate radius and strength. Look at the results of your analysis. Set the radius to match half * the size of blur circles or the full length of blur lines. Experiment. Sometimes different parts of the image require different treatments. Adjust the threshold in unsharp mask to reduce noise.
An example of Motion Blur
This is a pretty extreme example. I took this Euro in Broken Hill. I was panning at a low shutter speed in low light. Despite the obvious blurring there is something appealing in this image. I decided to try some sharpening in Photoshop (CS6). Looking closely at the ear I get a sense of lines going upwards and to the right. These are indicative that the Euro was moving in relation to the camera. I placed the cursor on both ends of a white highlight than had been elongated into a line. The Info panel told me that the line is 20 pixels long and is 57 degrees from horizontal. In the Smart sharpen filter I selected motion blur. I put in the angle 57 degrees and the radius 20 pixels. I chose 120%. In a split second Photoshop has produced a new image. There is more noise but it has also brought out details and textures that were only very faint in the initial image. I did some adjustment in curves and noise reduction and this is the resultant image:
My kids tell me it is still blurred. However to my eye it is much more presentable. In an abstract kind of way. At 20 pixels it is a bit of a shove. At 3-4 pixels however this filter can save an image and give it back some much wanted vitality.
Give it a try.
We have covered some of these concepts in previous editions of Camera Clips:
- Motion Blur – December 2013
- Noise – January 2012
- Sharpening – October 2010