Processing: Noise Reduction

There are many ways to reduce noise. You can use one or more software tools designed for the task. We'll stick to Photoshop, which gives us at least three ways to work: Filters, Carboni's Astronomy Tools Actions, and inverse masking.

Filter Method

This is launched using Filters > Noise > Reduce Noise... Choose the Basic mode, set Strength to 8, Preserve Details to 20, Reduce Color Noise to 80 if it's a color image and Sharpen Details to 0. You should play with the settings to get a feel for what works best for your image, then click OK.

Astronomy Tools

Because it's so easy to use this, let's use this to give you a little experience using layers.

  1. Open your image
  2. In the Layer Palette, drag the layer to the copy layer icon. Double-click the layer's name and change it to DSNR.
  3. With the DSNR layer selected run the Astronomy Tools action Deep Space Noise Reduction action. You'll find it listed in the Actions Palette. DSNR works only on the image's dark background; the similarly named Space Noise Reduction works on the entire image. For now we assume your image isn't full of nebulosity and our main goal is to reduce the background noise.
  4. Drag the DSNR layer to the copy layer icon to create a new copy. Rename it LCMF.
  5. With the new layer selected run the Astronomy Tools Less Crunchy More Fuzzy action.
  6. You can now see what the effects of the actions are by Alt-left mouse button clicking on the visibility icons for each layer. Clicking with the alt key makes the layer you're clicking on either the only layer visible or the only layer invisible. Play with this a little to get the hang of it. You should see that the DSNR layer has a significantly less grainy background than the original image. Comparing the outer portions of stars will show you the primary effect of Less Crunchy More Fuzzy: Stars are fuzzier in the LCMF layer than the DSNR layer.
  7. Make the LCMF layer the only visible layer, and then right click the it and select Flatten Image... Click OK to Discard Hidden Layers.
  8. Save your image.

If you're in a hurry, just do these steps:

  1. Open your image
  2. Run the Astronomy Tools action Deep Space Noise Reduction action.
  3. You can repeat step 2 as needed, but overuse will introduce artifacts into your image.
  4. Save the image

Inverse Mask

Deep Space Noise Reduction is an automated inverse mask method. For greater control over the process you can do it yourself.

  1. Open your image
  2. Copy it to the clipboard (Ctrl-A, Ctrl-C)
  3. Paste the clipboard to the image, creating a new layer (Ctrl-V). We'll call this the noise reduction (NR) layer
  4. At this point the NR layer should be selected. Add a mask to NR layer by clicking the tiny circle in a square icon at the bottom of the Layer Palette
  5. Select the mask by Alt-clicking on it.
  6. Paste the clipboard to it (Ctrl-V)
  7. Invert the Mask (Ctrl-I)
  8. Open the Levels tool (Ctrl-L) and adjust the black point toward the right to blacken the parts of the image you don't want to affect. Usually you'll adjust the black point upward to about the middle of the histogram and let the midpoint self-adjust along with it.
  9. Adjust the white point leftward to about the middle of the background peak (which is near the white end of the histogram)
  10. Click on the NR layer's thumbnail and apply a weak Gaussian blur to the mask (Filter > Blur > Gaussian Blur...) with radius between 0.5 and 2 that adequately smoothes the background. Don't worry about the blurring that takes place in your brighter target, as that will be masked out. Alternatively, you could use Filter > Noise > Reduce Noise... to smooth the background, or any other method you choose.
  11. Alt-click the NR layer's mask and apply the same smoothing you applied in the previous step.
  12. Toggle the NR layer's visibility to see if what you've done is satisfactory. If it is, merge the two layers and you're done.
  13. If the outer portions of stars look a little grainy, run Astronomy Tool's Less Crunchy more Fuzzy, perhaps in combination with a very weak Gaussian blur.


Table of contents Previous Page Next Page Last Page