|
Using
the Digital Darkroom to reduce noise/grain in images |
||
|
What you will learn and practice in this Seminar: Selecting image elements, and working in the Channels window to reduce unwanted noise. Noise/grain plagues high speed films (ISO 400 and higher) and low light, long exposure digital images. This seminar describes a technique for reducing this noise, and provides an example image of a sunset that one may use in applying these techniques as well as my final image file for comparison. You may want to print this seminar out and download the rough image file (P9230244-0.jpg) and examine the sky and clouds detail in Photoshop. Open up the Channels and look at the noise patterns in each of the Channels one at a time. You’ll see that each channel has a differing amount of digital noise that has the appearance of grain. In the case of this image it is the blue channel that has the most noise in it. Begin by selecting the sky only, with the magic wand tool, tolerance = 32, of the RGB image. Now in the Channels window select the Blue Channel and apply a Noise>Median Filter with a radius of 12 pixels, this will smooth out he noise in the blue channel. Go to the Green Channel and apply a median filter with a radius that will just remove the noise. Do the same with the Red Channel. Return to the RGB image and look at the effect on the sky. The sky in your image should now have a noise free appearance, and a smooth gradation of hue, darkening at the top of the image. The Blur Filters Leave more detail in the image than the Median filter does, and will be used to reduce the noise in the cloud patterns after they have been selected. Using the Magic Wand tool, set as before, select the darker clouds just below the sky. When you examine the Blue and Red Channels you will see that they contain the most noise. Starting with the Blue channel apply a Filter>Blur>Gaussian blur with a Radius of about 3.0 or so pixels. In the Red Channel apply a Gaussian blur with a radius of about 2.0 pixels. Proceed to the Green Channel and apply a Gaussian blur to that channel if you feel that the RGB image needs it.
At this point the work of reducing the noise in this image is complete. Practice with noisy images from your files. “Selecting” is an art, and the foundation of editing. Select only those elements of an image that have nearly flat colors (little or no detail) for noise reduction. If an image is busy with detail the noise will not be as readily apparent and may need but little channel adjustment. For more
information on Gordon Belgard-Leckenby check out: |
