Quick and easy noise reduction in Photoshop

This is an adaptation of a post I made on ShutterStock’s photographer support forum.

I see the question raised repeatedly of how to reduce noise in a photograph, and the prevailing wisdom is to use dedicated noise reduction software such as Noise Ninja or Neat Image. While both of these programs can do a fine job of cleaning sensor noise from an image, in most cases they are unnecessary — and a better, more natural-appearing result can be obtained by alternate methods. This is one such method.

Firstly, general points to remember when shooting, regarding noise:

1. Shoot with the lowest ISO possible for the situation. Aim for 100 for most things, 200 when necessary. Depending on your camera, you may be able to use 400 and even 800 without too much trouble, though they’ll probably call for more noise reduction in post.

2. Don’t underexpose. This is probably the most common source of noise. Sounds stupid simple, but it’s easy to miss.

Okay, you’ve made your shots and you’re home. Now on to the subject of fixing the noise that you couldn’t avoid when shooting.

This assumes the use of Photoshop.

1. After any color correction is done, if the image isn’t already in LAB color, I convert it now.

2. Examine the A and B channels individually for noise. In LAB, all of your color information and NONE of your contrast information is stored in these channels. This is important. If (when) I find noise in these channels…

3. Apply Surface blur (Gaussian blur is fine if you don’t have a new enough version of Photoshop to have Surface blur available) to the A and B channels. B usually requires a bit more than A. The strength of the blurring will depend on the needs of the individual photo, but you can go quite far without harming the final photo. You’re aiming to knock out ALL noise in solid areas of the A and B channels. Don’t worry if the “edges” in these channels start to look a little fuzzy.

This should remove most or all of the color noise from your photo (red, green or blue noise). It will have no effect on luminance noise (variations in brightness), but you’ll find that luminance noise is nearly always far less objectionable than color noise.

Bonus step: While you’re in LAB, this is the best place to do your sharpening if you desire. Select the L channel (remember, all of your contrast is in this channel alone) and apply Smart Sharpen (or Unsharp Mask if you don’t have CS2) to this channel. Start with an amount of about 100% with a radius of about 0.7 or so and adjust to taste. Threshold is usually set to around 2 to 4.

If you have CS2 and are using Smart Sharpen, turn on Advanced mode in the dialog. On the “Shadows” tab, set your Fade amount to about 50 or so, and your Tonal Width to about 50 also. This has the effect of reducing the sharpening effect in the darker parts of your image, where A: most of your noise lives and B: where sharpening usually has the least beneficial effect anyway. This prevents the sharpening from accentuating the noise in your shadows, while allowing you to really make the lighter areas of your photo pop.

Don’t forget to convert your image back to RGB color. Usually sRGB if you’re getting ready to submit them to ShutterStock. If Photoshop won’t let you save as a JPEG, it’s usually because you forgot to convert back out of LAB color. I forget this all the time.