Photoshop:
Within and beyond unsharp mask II.
Let’s continue from where we stopped the last time. Instead of theory, we are again strolling in the field of practice and will check out another set of actions for sharpening. We stopped at saying that it is better to avoid one-button and automatic solutions because using layers and color cannels yield a better result in sharpening photos. The advantage of layers lies in the possibility of fine-tuned sharpening, while channels can be distinguished and separated. Sharpening that has been carried out in Lightness channel in Lab color mode does not affect color channels and so the color noise does not increase after sharpening. But in this case it is more difficult to fight a significant luminance noise. Let’ clarify the two noises that often go hand in hand because you often meet them during photo editing and different methods are needed if you want to remove or decrease them.
In RGB color mode it is not easy to tell them apart when you look at luminance and color noise. As it was indicated earlier, red, green and blue color channels together account for brightness value. In Lab color mode, ‘L’(lightness channel contains lightness data, while ‘a’ and ‘b’ channels contain only the color data. The photo detail below (sky) shows all the noises without being sharpened. First the red, green and blue channels in RGB color more. The RGB window shows the whole of the noise content. You can see the amount the red (R), green (G), and blue (B) channel noise in the bottom half of the image. It is different in all three channels, but luminance and color noise both appear in all of them. If you carry out sharpening in only one channel, both components will increase.
The very same image detail looks different in Lab mode. Not so much if you look at the whole of the noise as if you look at each channel separately. You can only see lightness components in Lightness (L) Channel, while in channels ‘a’ and ‘b’, there’s only color noise. It’s recommended to carry out sharpening in L channel because that contains all the photo details too. Thus we only need to keep the increased lightness noise of L channel in check. You can forget about 'a' and 'b' color channels and increase of color noise too.
We have digressed a bit in the sea of noise, but increase in color noise is going to have an important role in the field of sharpening. The increase hurts the eyes especially in the homogenous parts of the photo which are poor in details, such as the sky, a plain house wall or a still water surface. The amount of noise increase is the same in detailed areas but its presence is less annoying because of the plethora of details, what’s more, it can even enhance the mood of the photo in some cases. Such instances are the foliage of a forest, sand or short animal fur. You see, noise is not always your enemy.
We will now sharpen an image so that the noise in the most noticeable, homogenous surfaces will not increase. A very useful tool for this is the so-called selective sharpening. To put it simply, we are going to do the sharpening along the necessary edges, excluding the homogenous areas from the process. This process requires a so-called contour mask, which blots out most of the smooth surfaces. Let’s get down to business.
We will carry out the sharpening in the following picture part. Based on the explanations above, the first step is to switch to Lab color mode. Apply Photoshop’s Image/Mode/Lab color menu command to do so. As you can see, there’s no change in the photo.
Only Channels palette on the right reflects the change. Instead of Red, Green and Blue channels, there are now Lightness, ‘a’ and ‘b’ channels. We need only Lightness from these, and not even the channel itself but its replica, since we would like to create a contour mask first. If we carried out these steps directly in channel 'L’, it would change the photo too. Our aim is to create a black-and-white photo where edges are clearly distinguishable. Create a replica of Lightness channel so that you can work on it with no harm. To do so, drag the headline of the channel onto the red icon where the red arrow points.
From here onwards this channel should be selected, named Lightness copy by the program. It is most simple to detect edges in Photoshop, because there’s a separate command for this. Choose Filter/Stylize/Find Edgescommand. The selected channel now resembles a pencil drawing: the edges are drawn, while the homogenous areas are left white. The stronger the edge, the bigger the difference in contrast from the neighboring areas, you get all the darker line as a result, and where there's no sufficient contrast difference, the line will be lighter or you get a white area. Little or no sharpening is required in these areas, as there’s nothing to sharpen there in fact. Only the noise would increase in the homogenous areas.
We have located the necessary edges, we now only have to soften the selected black contours and create an adequate mask with which to exclude the non-desirable areas. First make the edge contours stronger so that you can soften them. Choose Filter/Other/Minimum filter, then set 1 or 2 as radius value in the appearing window. The value 2 results quite firm edges, which is only recommended in case of very thin, somewhat uncharacteristic edges. A bigger radius value is required only in case of bigger-sized pictures.Radius 1 is generous enough in photos meant for the web (800 x 600 pixels). Since the result would yield quite a “heavy” mask, it is advisable to soften the contours so that sharpening later takes effect with soft transitions, providing a more natural result in the end. You can apply several filter commands separately or in a combination to soften the contour. Such examples areMedian or Gaussian Blur filters.
Let’s use both.
First, break the edges using Filter/Noise/Median command. Here it is best to give the same radius value as in the previous Minimum filter, which was 1(since we are speaking about a smaller picture). This will decrease the stain effect caused by Find Edges and Minimum filters, but for the really soft transitions you’ll also need Filter/Blur/Gaussian Blur filter as well. Radiusshould be the same or near value as used above. If there are a lot of dark, strong edges, you can use a bigger value too. In our case we chose the value1.
The result is more powerful than the effect of Find Edges, still it is softer, which makes it perfectly eligible for the role of a contour mask.
It is not appropriate for this role at its present stage because it’s now the inverse of what we need. If you change it to selection (mask), the dark areas will be blotted out and white areas are to be sharpened when we aim for just the opposite: we wish to sharpen along the dark edges and omit the now white parts. The solution is very simple: we have to invert the channels with the Image/Adjust/Invert menu command. This way the contour of the edges becomes white and the homogenous areas become black.
Selection is a simple thing from here. Press Ctrl and keep it pressed, then click on Lightness copy channel, on which you have been working so far. Only the white areas, that is the edges will be selected if you did it well. You don’t need Lightness copy channel any more so click on Lab channel inChannels palette. Our selection marker remains this time valid for the original picture. Therefore you can carry out the sharpening, which will only affect the edges. It’s a good idea to remove the contour of selection (contour mask) before starting sharpening. The mask remains and only the disturbing animated contour becomes invisible, but this way you can better track the effects of sharpening.
We are now returning to good old unsharp mask. ClickFilter/Sharpen/Unsharp Mask and sharpen the photo to your preferred extent. We don’t see changes in homogenous areas, but the selected edges will become more emphatic after sharpening.
The amount of noise will not increase significantly. Let’s see an example. The left part of the photo shows the state before sharpening. The medium section shows the state after the selective sharpening we just did. We applied quite a strong sharpening so that you can easily see that the noise in the homogenous background areas hardly increase but the contrast of the edges is significant. We did the same amount of sharpening in the right side image but without a contour mask and so it affected the whole picture. There’s an annoying amount of noise in the originally homogenous background while the contrast of the edges increased to the same amount as in the middle example.
With this, we have arrived at the next problematic field of sharpening illustrated best with the edges of marked contrast, the halo effect. This is the bright and dark edge contour arising from too much sharpening. In the next session, we will have a shot at eliminating this problem.
