Sharpening Tutorial - Part 2

Submitted by Bryan Wolf on Sat, 01/15/2011 - 9:37pm

Introduction

Welcome back! In Part 1 of this tutorial we covered Photoshop’s three most common forms of image sharpening. In this part, we’ll see how we can apply them to our imaging workflow. I’ll introduce a few methods of increasing mid-tone contrast, which helps our perception of sharpness and clarity. I’ll show a few ways to manage post sharpening’s arch nemesis, noise. Then, I’ll  briefly touch on Lightroom’s “Detail” pane, and finally wrap Part 2 up with a couple of real-world examples.

Midtone Contrast

A lack of contrast in the midtones can often be perceived as a lack of sharpness, or clarity. We can add some back with minimal fuss. We’ll use the High Pass filter for the steps below, but Unsharp Mask with a high radius achieves roughly the same effect.

1. Select the top most layer and “Merge Visible with copy”(Shift+Ctrl+Alt+E)

2. Set blend mode to “Soft Light” at 50% opacity using blending options from Figure 1a.

3. Filter > Other > High Pass ~50 pixels(anything from 40-80 usually works)

4. Adjust layer opacity to taste. I’ll usually use a value of 20-30%, but there are always exceptions.

Figure 1a - Midtone Contrast Blending Options (clck to enlarge)

Figure 1b -  Midtone Contrast: 50 pixels, Soft Light, 50% Opacity

The animated .gif above roughly shows the effect with a setting of 50% opacity. Remember this image has been compressed for web viewing and is only for illustration purposes. Don’t pass judgement until giving it a go for yourself! After applying some midtone contrast, you may find that your images will need less sharpening than you would have otherwise thought.

Noise Control

When it comes to sharpening, there is nothing more annoying than having your image turn into a snow blizzard because of noise. It looks horrible, and I don’t think I’m alone in thinking that. Don’t get me wrong, I shoot mostly film, I’m not afraid of grain. However, my goal is to return the digital file to it’s original qualities or make it better, without exaggerating that grain. I want the choice of graininess to be a film choice(or digital ISO choice), not a left over digital artifact. So how can we have our cake and eat it too?

High Pass Myth

In theory, High Pass sharpening should be better at making the important bits sharp and keeping smooth tones smooth. In reality, it’s not much better than Unsharp Mask or Smart Sharpen in this regard. I’ll show you:

Figure 2a - Base image with noise (clck to enlarge)

In figure 2a above, we have our gradient image from part 1. I’ve then added some digital noise on top of it using an overlay layer. This will be our test image for the next few illustrations. Keep in mind these images have been reduced from their original size of 4”x1” @360PPI. Full resolution versions will be available in part 3 for you to print yourself and the differences will be easier to see.

In figure 2b below, I have applied the same sharpening settings we used in part 1. 250% for Unsharp Mask and Smart Sharpen's “Amount”, 100% opacity “Overlay” for the High Pass layer, and 3 pixels as the radius for them all. Compared side by side, the High Pass looks less gritty than the other two, but it also looks like it has not been sharpened as much either.

Figure 2b - Sharpening noise (clck to enlarge)

I suspect that the majority of people who claim that High Pass is much better for masking grain, did not use visually comparable settings. Below is all three set to what I feel is more comparable visual sharpness.

Figure 2c - Sharpening noise with comparable settings (clck to enlarge)

The High Pass is still slightly better at very high magnification, but not enough to give much concern, particularly at print DPI. In my opinion, there are better ways to control noise artifacts than the method of sharpening you use.

Masking for Grain Reduction

A far more effective way to manage grain is through both noise reduction and masking. Reduce Grain? That’s blasphemy for a film shooter, right? Maybe. I like grain, but the digital process has a way of exaggerating it. There is aliasing at the scanning stage, then sharpening artifacts, and then interpolation artifacts, not to mention dithering if you print with an inkjet. Digital prints from film often end up having more visible grain/noise than a darkroom print from the same negative. I don’t think of noise reduction as removing the character of the film, but managing digital artifacts. The decision to use noise reduction needs to be made on a shot by shot basis, but the masking techniques covered below are useful for every image we intend to sharpen.

Creating the Mask

Ideally, we want a mask that protects the edges from noise reduction, but allows it in areas of smooth tonality.  Photoshop’s “Find Edges” filter, plus a few tweaks, fits the bill. Here are the steps to create an edge mask:

1. Image > Calculations. I prefer to add some contrast before masking, this will help Photoshop find the edges during a later step. We do this using “Calculations” and the “Overlay” blend modes with the following settings.

   

   Figure 3a - Calculations (clck to enlarge)
    

2. Select the new channel(in the “Channels” palette) and rename it to “EdgeMask”

   
    

3. Filter > Blur > Gaussian Blur.  Slightly blurs the mask to soften the noise/grain. The amount is up to you, but I suggest a smaller radius for fine grained, high detail landscapes, and a larger radius for portraits.

   

   Figure 3b - Blurring the channel (clck to enlarge)
    

4. Filter > Stylize > Find Edges. Creates a grayscale mask of the edges.

   

   Figure 3c - Find Edges (clck to enlarge)
    

5. Filter > Blur > Gaussian Blur. Smooths the transitions between white and black; a falloff.

   

   Figure 3d - Blurring the mask (clck to enlarge)
    

6. Image > Adjustments> Curves. Adds back some contrast lost from the previous step.

   

   Figure 3e - Adding contrast to the mask (clck to enlarge)

Noise Reduction Layer(s)

Now that we have created our edge mask we need to apply it. But first I’ll run some noise reduction. I use Noise Ninja, but you can use any noise reduction method you want. Just make sure and turn off all sharpening in the noise reduction dialog(if it has it). I like to separate my noise reduction into two layers, one that is set to luminosity blend mode, and one that is set to color blend mode. I do this so that I can dial the amount of noise reduction separately for both color and luminosity by adjusting their opacity. Grayscale images obviously only need one layer due to the lack of color. Here is a layer setup for both a color and grayscale image:

Figure 4 - Noise reduction blend modes (clck to enlarge)

Applying the Edge Mask

Now we are ready to apply the layer mask to the noise reduction layers. To do this, we’ll use a couple Photoshop shortcuts. These are the steps:

1. Select the “Channels” palette and “Ctrl” click the “EdgeMask” channel. This selects it.

2. Select the “Layers” palette and then one of the noise reduction layers.

3. Then click “Add layer mask” at the bottom of the layers palette.

4. Repeat for any noise reduction layers.

5. Verify that the noise reduction layers are masked.

Figure 5 - Applying the edge mask (clck to enlarge)

Masking for Sharpening

Now that we have an existing edge mask, this part will be very easy. First you create your sharpening layer by selecting your top most layer and “Merge Layers with copy”(Shift+Ctrl+Alt+E). Remember to apply your “Blending Options”, outlined in Part 1, Then follow these step(s):

1. “Alt+Shift Drag” your mask from the noise reduction layer to the sharpening layer.

That’s it! I told you it would be easy. “Alt+ Drag” copies your mask, “Shift” inverts it. We do it in one step to simplify. This mask can be used on any of your sharpening layers. and modified by painting white or black into areas you want more or less sharpening. You can also tweak the mask by using levels, etc. We used the opposite of the Noise reduction mask for obvious reasons. We wanted to protect the edges from noise reduction and we wanted to protect smooth tonality from sharpening; our edge mask can do both.

Putting It All Together

Figure 6 - Layer stack (clck to enlarge)

A finished image prepped for printing, might have a layer stack something like figure 6(above).With all the separate layers you may run into memory problems, especially with scans of medium and large format. I can not tell you how to best deal with these other than to suggest an alternative workflow. You may want to save a flattened image at each point in the process. For example, save your original scan, then save a version with noise reduction, then flatten that and make a version with your tonal edits, etc. You’d end up with approximately the same total file size on disk, but you will not run into memory and swap problems when opening the individual edits. With a powerful machine and a 64-Bit version of Photoshop, I don’t think you would have much of a problem with any reasonable sized image.

Photoshop Lightroom

Up until now, we’ve been discussing a pure Photoshop workflow. Some users prefer to combine the organizing and simplified work-flow of Lightroom with the power of Photoshop when needed, myself included. I will typically bring a scan into Photoshop for initial tweaks, noise reduction(if needed), and a round of capture sharpening. From there I will import the image into Lightroom to perform simple tonal edits and also use the “Detail” panel for my “content sharpening”. In the rare event that I can’t accomplish what I want there, I’ll round trip back to Photoshop. Lightroom’s “Detail” panel and “Clarity” slider offer much of the previously outlined methods in a simpler package. They don’t offer the same control as manually doing everything in Photoshop, but for the majority of images, it’s more than good enough. I’ll briefly discuss the “Clarity” slider and “Detail” panel below.

The “Clarity” Slider

To the best of my knowledge, the clarity slider is very similar to using the High Pass filter for midtone contrast as illustrated at the beginning of this tutorial.

Figure 7a - Lightroom Clarity (clck to enlarge)

The “Detail” Panel

The detail panel wraps many of the techniques discussed earlier into one tight packge. However, In typical Lightroom fashion, it’s simplification gives up ultimate control. You can achieve say... 90% of what you can in Photoshop, but should you want more control over what gets masked, etc. you’ll need to make a trip to Adobe’s more powerful editor. However, In most cases you’ll find that Lightroom offers enough options, and you’ll appreciate it’s speed and simplicity.

I did not work with Adobe to build Lightroom, so I have no way of knowing what the software is actually doing, take what I say about these options as opinions.
 

The "Detail" panel has a pretty handy trick. If you hold the “Alt” key while dragging any of the sharpening sliders, you see a grayscale image of the actual effect being applied. I’ll illustrate this below.

Figure 7b - Lightroom Detail Amount (clck to enlarge)

Figure 7b - Lightroom Detail Radius (clck to enlarge)

Figure 7b - Lightroom Detail (clck to enlarge)

Figure 7b - Lightroom Detail Masking (clck to enlarge)

The Result:
 

Figure 7f - Lightroom Detail before and after (clck to enlarge)

If I had to make a guess, I would say the amount slider works like an amount slider... no surprises there! : ) The radius appears to be controlling a High Pass radius. The Detail slider looks to be adding traditional Unsharp Mask back onto the High Pass result, and the “Masking” slider is making a mask similar to the Photoshop way, but in real-time. Very handy. These are just guesses though, and I wouldn’t be surprised to find out it is quite a bit more complicated under the hood... or I'm flat out wrong. With that said, it seems that the “Detail” panel combines many of the same tricks we use in Photoshop. This is not so surprising, considering Bruce Fraser and his colleagues are said to have been instrumental in getting these implemented in Lightroom.
 

I’ve seen it mentioned that the Noise Reduction sliders are also linked in some way to the Sharpening settings. I can not verify this and because I prefer Noise Ninja, I have chosen not to visit it in this tutorial.

Real World Examples

Example 1

First I’ll work on a snapshot of my daughter. The image was shot on 35mm HP5+ and developed with divided Pyrocat MC. I scanned at 2540SPI and downsampled to 5"x7"@360PPI using “Bicubic” to avoid any sharpening. I chose this shot for illustration because HP5+ is relatively grainy emulsion, and I wanted to see if we could sharpen without introducing artifacts.

Figure 8a - Real world 1; Base image (clck to enlarge)

These are the steps I took to get the final result in figure 8b below. Please note that I was trying to keep my edits very subtle; they may be hard to see in the final result. You’ll have to take my word, that the differences are more visible when in Photoshop.

Steps

1. Created an edge mask, but modified it slighty. I wanted less grain reduction in the Iris of her eye, so I painted some black into the mask in those areas.

2. Created a grain reduction layer with an opacity of 66% and applied the edge mask.

3. Added midtone contrast using High Pass at 50 Pixels, blending options from Part 1 of the tutorial, and set the layer to 21% opacity “Soft Light”.

4. Added a bit of contrast using curves.

5. Created a capture sharpening layer above my tonal edits, to minimize artifacts. Smart Sharpen with Amount: 80, Radius: .3, lens blur, more accurate. Set layer blend mode to "Normal" with 80% opacity using the "Blending Options" from Part 1. I then applied the inverted edge mask.

6. Created a content sharpening layer using High Pass at .8 pixels, blending options from Part 1 of the tutorial, and set the layer to 33% opacity “Soft Light”. Created an all black mask and painted white just around the eye and brow.

7. Created an output levels layer to adjust for printing gamma and black and white points.

Figure 8b - Real world 1; Sharpened image (clck to enlarge)

Figure 8c - Real world 1; Before and after (clck to enlarge)

Example 2

This image was captured on 35mm Fuji Reala, scanned at 3175SPI, and downsampled to 5"x7"@360PPI using “Bicubic” to avoid any sharpening. I wanted to covert this to a high-contrast black and white  image with a slightly warm tone and good sharpness. Below you’ll find the original color image and the steps I took to get the final result(figure 9b).

Figure 9a - Real world 2: Base image (clck to enlarge)

Steps

1. Created an edge mask

2. Created a "color" noise reduction layer with an opacity of 80% and applied the edge mask.

3. Created a "luminosity" noise reduction layer with an opacity of 33% and applied the edge mask.

4. Added midtone contrast using High Pass at 50 Pixels, blending options from Part 1 of the tutorial, and set the layer to 66% opacity “Soft Light”.

5. Converted to B&W, a bit of contrast using curves, and a warmtone using a seperate curves layer.

6. Created a capture sharpening layer above my tonal edits, to minimize artifacts. Smart Sharpen with Amount: 85, Radius: .5, lens blur, more accurate. Set the layer to  "Luminosity" blend mode at 100% opacity using the "Blending Options" from Part 1. I then applied the inverted edge mask, but adjusted output levels to 85, 255  so that there was no values lower than 85 in the mask. I did this because I wanted sharpening across the whole image, but less in the smooth tones.

7. Created a content sharpening layer using High Pass at .5 pixels, using blending options from Part 1 of the tutorial, and set the layer to 66% opacity “Hard Light”. I then copied the same modified edge mask from the “CaptureSharpening” layer.

8. Created an output levels layer to adjust for printing gamma and black and white points.

Figure 9b - Real world 2: Edited image (clck to enlarge)

Figure 9c - Real world 2: Before and after (clck to enlarge)

Conclusion

So there you have it. We’ve covered quite a bit in this part of the tutorial. I hope that you are able to try some of these techniques in your workflow and find the results to your liking. I can’t begin to tell you how difficult it is to show results on screen with such small images, not to mention web compression. Good news is, in Part 3 of the tutorial I’ll be printing these images, scanning in the prints and making some observations of the results. I’ll also explain print sharpening, wrap it up with some Photoshop actions, and give you an image to download and print for yourself. I hope you’ve found this tutorial helpful, and as always, please leave me a comment to let me know how you got along. If you have any questions plaese feel free to ask them inthe comments below or send me a email using the contact form.

Regards,

Bryan

P.S. I know this was a long article with lots of scrolling and I really appreciate that you made it this far! I'm looking into ways to break the content up into pages, so there will be less scrolling, but  It's not quite ready for this tutorial. Sorry 'bout that. : )

<- Back to Part 1

• Login to post comments