Sharpening - A brief introduction

Sharpening

- A brief introduction -

by Micholand



Sharpening - what's that, how does it work and why is it necessary?

Image sharpness is always something which is on photographers' mind and a lot of effort is put into. Chasing the perfect piece of glass that produces images with high contrast and resolution and therefore the question how good and sharp lens 'abc' is attached to camera 'xyz' is discussed frequently. Not to forget the general question-and-answer game what exact settings to use to sharpen an image so it jumps right out at you and also when to do this in the editing process. Of course there is no simple answer to all that, but it's commonly accepted that sharpening is needed as the human visual system depends to a great degree on recognizing details by detecting transitions (edges) which our eyes pass as 'the information' to our brain.

Sharpening is a large topic and there can be found tons of information in books, magazines and step-by-step tutorials of different methods, but in the end there's one important general rule and that's understanding how sharpening works. Otherwise it easily can go wrong - from too much to not enough.
In a nutshell sharpening is enhancing the edge transitions of an image, respectively increasing the contrast around edges. That's all!

Nearly all digital pictures are to a certain degree fuzzy which has nothing to do with the eye behind the viewfinder pressing the shutter or a lemon lens, but rather how the image is caught by the sensor. Photons with a continuous tone and colour gradation in real world are converted into discrete pixels in the digital image. Due to the fact that the sensor has a certain kind of resolution available, only a fixed grid of pixels can be sampled as well as colours detected via the used colour filter array (CFA). Therefore in-between values, especially for diagonal lines, have to be interpolated by the use of demosaicing algorithms which at the same time introduce some loss in sharpness. Furthermore most digital cameras employ an anti-aliasing filter - a low-pass filter to block short-wave light portions - over the sensor to not produce, respectively lessen, coloured artefacts and moire patterns the digital sampling normally tends to create and which are hard to remove. Since the light is blurred slightly by this filter there is a further tendency to make edges less distinct.
So sharpening is needed to a certain degree to enhance those edges of a digital image again after the capture process.

Edges in images always involve darker tonal values adjacent to lighter ones. Now locating such pixel areas of contrast and amplifying their changeovers, by lightening the light tonal values and darkening the dark tonal values at this joint, the transition is emphasized further as the contrast around these edges is increased. When this is done a so called halo - an outline that sharply separates an area, normally having the inverse colour along the outside edges of that object - is created and that makes the transitions more visible and hence the entire image seems to be sharper.
So sharpening works by increasing the contrast around edges. Even though sharpening is closely related to contrast, simply increasing the contrast over an entire image doesn't improve its sharpness. It just produces an over contrasty image. Instead, successful sharpening demands to localize those parts in an image that actually represent edges and then to boost this local contrast.


Sharpening - its limitations

Understanding how sharpening works also leads to an understanding of its limitations. Sharpening can improve a good image considerably, but trying to correct a fuzzy and out-of-focus image into a sharp one is beyond what it is intended for. There already must be some fine contrast transitions in the original shot so that the sharpening filter, with some given threshold, can look for them. If there are no such differences between dark and light - that is named contrast - nothing will be done and thus there is no visible effect. Therefore already the original captured image must inherent as much contrast as possible or in other words sharp edges to respond easily to the sharpening tools. In short, sharpening can't fix sloppy focus or even insufficient depth of field.


Sharpening - when to sharpen?

Often it is claimed that sharpening shouldn't be done repeatedly and that it's the last step in the image editing process, after any colour correction, levels and curves adjustment or edits involving cloning, masks, et cetera, otherwise it will likely end up in unnatural looking results, e.g. with sharpening halos along high contrast edges. To a certain degree this is all logical and valid, but on the other hand only half the truth. Of course all major changes done to the image like colour correction, saturation changes, distortion corrections, et cetera should be performed before the final sharpening. Thus sharpening will be more effective and simplified if the image contrast already is optimized as the sharpening filter boosts contrast along edges, but also the overall image contrast contributes to what looks like sharpness.
But sometimes not all things can be addressed in one single round of sharpening. On the one hand bring out fine details and locally increase the contrast along edges, then again sharpen for the output process which may introduce additional specific softness. In the end this may lead to an insufficiently sharp result or it is overdone and grossly over sharpened. So a one-pass sharpening approach often doesn't work that good. Otherwise doing it multiple times without taking into account a few things will easily result in over sharpened images.
The pitfalls mainly are:
    - Make the image look perfect sharp on screen, for example during RAW conversion, then sharpen once again at a later stage for final output - often this gives poor and nasty results
    - First sharpening done to the whole image instead of using a mask that isolates edges, for this reason noise and flat textured areas are sharpened, too. With the second run the image becomes over sharpened then.
    - Applying the wrong kind of sharpening for the image, so unwanted detail or noise may be exaggerated. A busy image with lots of fine details demands a different kind of sharpening than an image with soft wide edges.

In principle sharpening - if done carefully - can be applied at various stages of the post-processing if headroom for the final output sharpening is still present.

In general sharpening can be split into these three areas:

    - Capture sharpening
        Applied early in the editing process to restore any sharpness lost in the capture process itself, whether scanned film or direct digital capture. Here the image detail is of interest and needs to be accentuated without exaggerating the noise, hence must be done carefully!
    - Creative sharpening
        Applied to accentuate specific things in an image, e.g. eyes in a portrait shot; isn't required for every image and also is highly subjective
    and finally
    - Output sharpening
        Applied to images that have been sized to the final output resolution and to counteract the introduced softness of the output process; of all three, this is by far the most relevant and always should be performed

The amount of sharpening is firstly dictated by the image content itself and secondly how an image will be used finally, means what output process for the image is chosen. An image posted to the web or viewed on the monitor asks for a different type of sharpness than one which will be printed, whereas also here an output to an ink-jet printer needs another level of sharpening than a dye-sublimation printer. So the point is, when you’ve finished all your other edits save and backup this version and then use it as the master to generate versions sharpened for the specific image output process.
In addition to that there is no fix formula that can be applied to all the photographs, each image requires some work to determine the adequate sharpening parameters - a landscape shot needs different treatment than a portrait shot of a model.



Edited by Micholand - 20 February 2010 at 16:24

Advanced sharpening techniques

Sharpening via LAB Lightness Channel

LAB colour mode is composed of three channels, the "lightness" channel, with most of the tonal values of the image in it, and two colour channels, "a" which contains red/green information and "b" containing blue/yellow info.
The lightness channel in LAB mode contains the brightness information of the image, but not any noise caused by the colour channels. So isolating the lightness channel avoids sharpening the colours of an image, meaning only the black and white brightness and shadow values are affected and therefore any possible colour fringing and artefacts that can occur when sharpening the whole image in RGB mode are avoided.

    1. Convert the image from RGB to LAB colour mode (Image -> Mode -> Lab Color)
    2. In the 'channels palette' select "Lightness" channel only;
        Image will show up in b&w now
    3. Sharpen the "Lightness" channel via "Unsharp Mask" filter (Filter -> Sharpen -> Unsharp Mask)
    4. Return to the standard LAB channel view (-> in the 'channel palette' select "Lab" channel)
    5. Convert the image back to RGB colour mode (Image -> Mode -> RGB Color)


Sharpening using Highpass Filter

A good alternative to sharpen images is to use a technique known as highpass sharpening. The "Highpass" filter is capable to detect edge details within a specified radius by isolating high contrast image areas from their low contrast counterparts and therefore provides the essential and important data for sharpening.
Yet another advantage compared to the traditional "Unsharp Mask" filter is that the changes made aren't directly to the images itself, instead an adjustment layer is used. So the original image is left untouched and there is still the possibility to alter the amount of sharpening later on, either overall or selectively. Also the sharpening step can be repeated based on the original image data using the "Highpass" filter with different "Radius" settings.

The workflow which takes advantage of highpass filtering looks like follows:

    1. Duplicate your image layer
    2. Select the duplicate layer and apply the "Highpass" filter (Filter -> Other -> High Pass) to this new layer; in the "Highpass" filter dialogue adjust the "Radius" setting;
        A low radius of 1.0-2.0 extracts extremely fine details whereas a higher value (5-10) extracts coarse information, the exact value isn't that critical since in the final step things still can be fine tuned - but normally values higher than 10 aren't that effective any more.
    3. Change the 'blending mode' of the highpass filtered layer to either "Overlay" or "Soft Light" or "Hard Light"
    4. Then adjust the final sharpness using the layer's "Opacity" slider

Highpass sharpening may not produce better results than "USM", but it does let you sharpen on a layer without altering the original image data.


Sharpening using an Edge Mask

Instead of globally sharpening the image, this technique sharpens the edges of the main subject without increasing image noise which is especially helpful when sharpening grainy images. This will give a more smooth result and moreover avoid a speckled output compared to "Unsharp Mask" filter, even if the threshold value in the "USM" dialogue is used.
In short, first a mask is build containing just a selection of the essential edges in the image, then sharpening is applied to just these areas under the mask.

Following set of steps need to be done:

    1. Go to the 'channels palette' and duplicate one of the colour channels;
        'Red channel' most of the time works well as it has good b&w contrast, especially along edges
    2. Select the new duplicate channel and apply the "Find Edges" filter (Filter -> Stylize -> Find Edges);
        All edges will now show up in black and background in white
   
The following step is optional and only makes the edge mask effect more apparent:
    2b. Make a 'levels' adjustment (Image -> Adjust -> Levels) of the mask and adjust the black+white point sliders so that the blacks are really black and the whites are really white in this mask, also move the grey point slider to the left near the black point slider
   
    3. 'Invert' the light and dark tones in the edge mask (Image -> Adjustments -> Invert);
        "Find Edges" produces black lines against white background, but to select the edges for the sharpening it has to be the other way round!
    4. Now slightly thicken the edges found by expanding the white areas (Filter -> Other -> Maximum; and a Radius of 2 pixels)
    5. Soften the edges around the edge mask applying a slight 'Gaussian blur' (Filter -> Blur -> Gaussian Blur; and a Radius of 4 pixels) so that there is a smooth transition
    6. Convert the edge mask to a 'selection' outline (Select -> Load Selection; make sure the edge mask appears as the channel in the dialogue)
    7. Return to the standard RGB channel view;
        Now the "marching ants" should outline each edge area
    8. Finally sharpen this edge selection via "Unsharp Mask" filter (Filter -> Sharpen -> Unsharp Mask)
    9. Delete the 'selection' finally


Unsharp Masking via image calculation

Here's another advanced method using Photoshop's "Apply Image" command to sharpen an image via unsharp masking.
It's somehow replicating an analogue wet darkroom technique, where on the basis of a b&w negative a slightly blurred positive film is duplicated, then both are placed on top of each other and a print is created. When the enlarger is focused on the bottom negative, the top and out-of-focus copy creates a contrast mask that boosts the contrast along the edges in the image as the out-of-focus dark contour burns the dark side of the edges and the out-of-focus light contour dodges the light side of the edges. Therefore the 'unsharp mask' has the effect of increasing the apparent sharpness in the final print.

Summary of the steps:

    1. Duplicate your image layer twice, so that you have two identical copies of it
    2. Select the topmost (third) layer and blur this copy by applying 'Gaussian blur' (Filter -> Blur -> Gaussian Blur; and a small Radius, e.g. 4 pixels)
    3. Select the second layer (1st copy)
    4. Go to 'Image -> Apply Image', then just alter the following settings in the dialogue
        a. Choose as the 'layer source', the previously blurred (and topmost) layer
        b. Choose as 'blending option' "Subtract"
        c. 'Offset' value is set to "128"
        d. Apply the settings via the OK button
    5. Select the original image layer
    6. Again go to 'Image -> Apply Image', now use the following settings in the dialogue
        a. Choose as the 'layer source', the just modified second layer
        b. Choose as 'blending option' "Add"
        c. 'Offset' value is set to "-128"
        d. Apply the settings via the OK button
    7. The original image is now sharpened and the two duplicate layers aren't needed anymore and can therefore be delete


Third-party Tools

There are a number of third-party tools that automate or augment the sharpening process, like Pixel Genius's PhotoKit Sharpener or Nik Software's Sharpener Pro. Of course they are not free, but cost some money; however both offer free download trial editions.
If there really is a need for such third-part tools anyone needs to know for himself, though.




I hope this little article is somehow helpful and sheds some light on the topic sharpening. Of course it's only an abstract of some basic methods to sharpen images and for sure doesn't cover all methods, since sharpening is probably a subject where a lot has been written about. Hence one will find many various books, magazines and online-tutorials to look into this in much more detail and find further informations about advanced sharpening approaches.
Last but not least, please feel free to comment and also post any other advanced sharpening method which hasn't been covered yet in the above sections, for the sake of completeness.

Thanks for reading!
Michael

Thanks a LOT Micholand! Very helpfull!

Very informative write-up Micholand, Vielen Dank!

You might perhaps add in the beginning of the second part that you are specifically reviewing the sharpening options in Photoshop. Most of the relevant menus and functions are labeled or laid out a bit differently in other common editors like GIMP and PSP.

What an excellent post here - I will certainly be bookmarking this and giving a number of options a real try out. It doesn't hurt to have a range of options open to you depending on the nature of the image you want to sharpen.

A big thank you to Micholand - this is a significant amount of work here.

Kind regards, Neil.

Wonderful job

BRAVO! Thanks for the great explaination on "why sharpening" and the various sharpening tools.

LaterZ!
Darren!!