Grain is bad/Grain is good, Part I

This is part one of a two part series on working with grain.

When I say grain, I mean digital noise. It makes more sense to me creatively to relate the two concepts of ‘film grain’ and ‘digital noise’. Digital noise is nothing more than more-orderly film grain, and I’m not sure why it has the bad reputation that it does. So, moving on.

First, let’s talk about what grain is, so we can better understand how to get rid of it. Know Thine Enemy!

Here’s an extremely detailed article on digital noise. Read up, and then come back. We’ll talk about what we can do with and about grain. I think of grain as the camera’s guesswork. The higher the ISO, the more guesswork the camera is doing on the detail it is being asked to record accurately. More grain existing in an image represents the camera’s estimations being further and further from ‘accurate’. It makes guesses based on color and luminance (which is like brightness). So, inconsistencies of both brightness and color can exist. In this post, you’ll learn how to get rid of, de-emphasize and mitigate graininess in your image.

Most of my grainy images come from shooting concerts, so we’ll start with one of those.

Malevolent Creation at masquerade Atlanta

This is a shot of Malevolent Creation, a metal band I shot at the Masquerade in Atlanta. I shot this with my Canon 30d at 3200iso, 1/60th of a second, at f/1.8, with my 50mm f/1.8 lens, without a flash. I pushed the exposure up. This image is extremely grainy, as you can see in the 100% zoom below.

100% zoom noise detail

Notice in this next image how, even the sharpest areas of the image are suffering from the noise’s ‘averaging’ effect. Yikes! Let’s get started.

Before we start attacking the detail of the image, let’s start with some global adjustments. There is a lot of noise in the darkest areas of this image. Let’s push all those noisy shadows even darker so they get knocked out and less noticeable. I used a curves layer, like so:

curves adjustment layer

Next, we’re going to attack the noise more directly by looking at the image’s individual color channels. If you’re using photoshop’s default window setup, it’s right here:

Channels pallete

Let’s take a look at the individual channels, looking for noise. Here’s Red, Green and Blue, respectively:

Red Channel

Green Channel

The blue channel is a complete asshole

As a rule, the blue channel is a complete dick about noise. The red channel is bad in this case, but not always. The green channel is fine, as it usually is (this, I believe, is because there are twice as many ‘green’ sensors in your camera’s sensor as either other channel). The blue channel is almost always the noisiest.

I’ll knock out some color noise by reducing the saturation on the image a little bit. I’ll use hue/saturation for this.This is a good idea because it not only reduces color noise, it also helps control the extra-saturatedness of high ISO images. Check out the shadow area on this dude’s arm while he throws up the devil horns.

saturated noise is more evident

and after:

desaturated noise is less evident

Much better!

So now that we know where we need to reduce noise, Photoshop has a reduce noise filter that we’re going to use next (there are other 3rd party tools, like the highly-lauded Noise Ninja, which are considered by some to be better). So, I’ve created a new merged layer (I keep all my previous layers underneath, so I can preserve the ability to edit and redo my work later).The reduce noise filter has ‘advanced options’, which let you work on noise on a per-channel basis. So, I’m going to make my decisions here based on my observations earlier, preserving more detail in the green channel, and really obliterating the blue channel.

‘Preserve detail’ is a threshold control, just like in Unsharp Mask. It’s Photoshop’s way of figuring what is noise, which should be destroyed, and detail, which should be retained. Strength controls how much smoothing (or whatever it does, exactly) it will apply to the areas that are ‘noise’ and not ‘detail’. I prefer to set it higher overall than I think I should, and then paint in the noise reduction with a layer mask. Noise reduction can be hell on the actual detail you want to retain, because the threshold control is not perfect.

At this point, I’m done working on noise, and I can put some saturation back in, since I’ve gotten rid of the noise that would have caused a problem. At this point, I’ll make my creative decisions about the image: toning, contrast, saturation or what-have-you.

Lastly, I’m going to create a new merged layer again, and run Unsharp Mask. It’s important to get rid of noise before sharpening your images. Photoshop can’t always tell the difference between noise and detail, so Unsharp Mask can end up intensifying grain. I am going to use a higher-than-usual radius setting, because the noise in the image ‘scattered’ the detail. A higher radius will allow me to ‘corral’ it back into defined edges.

Here’s the before and after:

before noise reduction After noise reduction

Much better. Great! Next time, we’ll talk about getting convincing film grain in our images. See you then!


  1. Awesome weblog! Came upon it by accident when searching for the Louis Vuitton SS08 making-of video. Great breakdown and a great site. Keep it up, I’ll be checking back =)

  1. 1 Saturday Links Fever [2008-03-01] at All Day I Dream About Photography

    [...] Grain is bad/Grain is good, Part I Clique A photoshop trick to remove noise from your digital picture using Adobe Photoshop [...]




Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s



Follow

Get every new post delivered to your Inbox.

%d bloggers like this: