This is my actual “normal” workflow for noise reduction and sharpening for most of my images using Adobe Photoshop (CS or higher). The tutorial is quite long so I’ll split this into parts.
While there are many plug-ins available commercially to deal with noise and sharpening, it is always good to know how to get the results you need with the baseline tools that can do as good of a job than the commercial products.
My workflow is a combination of Bruce Fraser’s AWESOME three-pass sharpening technique, plus Michael Kieran’s inputs on how the tools involved can alter local color contrast which affects how viewers interpret perceived noise and sharpness.
Before we begin, I must mention that this is NOT a tutorial on how to salvage under-exposed, high-ISO images. You can use the best DSLR in the market and it won’t matter if you severely blotched your exposure during capture. Your image will still look smudgy and hard-grained with little to no midtones whatsoever.
It is also NOT a tutorial on how to sharpen blurry images due to slow shutter speed or camera shake. This is a tutorial on how to make your good images even better for your viewers.
The image below is a snapshot of my son just as he was heading for bed. It was shot with a 10MP DSLR with a 50mm and on-camera flash at ISO 800. The posted image is how it looked during capture, no adjustments were made on the exposure, just minor white balance adjustments.
Here’s the zoomed crop. We’ll primarily concentrate on the shadow and midtone details for this section.
As always, we want to have a starting and back-up image to work with, so we duplicate the Background layer.
I want to tackle the noise issue first. Shot at ISO 800, there are some visible chroma noise in the image, particularly in the shadows. There are many good noise-reduction software plug-ins available, but as I’ve mentioned earlier, I want to stick to the default tools and filters available in Photoshop for my workflow.
I’m analyzing the noise pattern and frequency through the Channels palette. Inspect each color channel and see which channel contains the most noise.
Here you can see the Red channel.
the Green channel,
and the Blue channel.
We can clearly see that the Blue channel contains the most grain and noise pattern around the shadow areas, so I’m going to tackle this layer first by using the Median filter (FILTERS >> NOISE >> MEDIAN). The Median filter tackles luminance noise quite well, and if you’re applying it on individual color channels, you can selectively remove certain color noise more efficiently than doing a one-pass Reduce Noise filter on the RGB image.
The amount used depends on how noisy the image is. I prefer to start with a low Radius amount and just re-apply the filter as necessary. In this case, I applied three instances of Median filter at 1px Radius setting.
I applied one instance on both the Red and Green channels as well. Again, the amount of noise-reduction used depends on the image.
Mouseover the image to see the “before NR” state of the image, particularly the darker areas around the iris.
Remember to change the value of the Median filter if you need more noise reduction. If you reduce more noise and grain, you’ll need more sharpening as well to get details back to reasonable sharpness. It’s prudent to determine how much noise you want to eliminate as the more noise is reduced, the less details are retained. Your subjective image analysis will determine the happy balance between noise and detail retention.
I need to reiterate this – EXPOSE CORRECTLY to begin with to minimize noise regardless of ISO settings.
We’re done reducing the noise for this image. Let’s proceed to the sharpening phase of the workflow.