My Lightroom Workflow – pt. 7 – Noise and Mimicry

details.jpgAfter showing you how for high-ISO shots today’s cameras rely more and more on noise reduction, no matter if it’s done in-camera, or in post processing by the camera manufacturer’s RAW development software such as Capture NX or DPP and after explaining how to save ISO-dependent development settings in Lightroom, one question is still open:

How do we make Lightroom mimic as close as possible what the manufacturers do in that respect?

Or even more important, let’s first have a look at what they actually do when reducing noise. I don’t have in-depth knowledge of any of the camera manufacturer’s specifics or how Lightroom exactly does it, but based on what I’m seeing with my own eyes and based on my knowledge about noise reduction, the least I can do is try a sophisticated guess.

Noise comes in all sorts of different incarnations, depending on many factors, for example it depends on what color model you work in. If you’re a photographer, you most likely live within the realm of the RGB channels. Red, green and blue. What you end up seeing in a final image is always a combination of those three channels. Whereas a picture of a neutral grey surface will have more or less the same amount of red, green and blue, a pixel taken out of a picture of a yellow flower will have a lot of red and green and almost no blue. So you can imagine a photo to actually be three photos, one for each channel.

The mixture of those channels will also factor into the noise you see in a final picture, as each channels has its characteristic noise profile. Very often you’ll find that the blue and green channels have much more noise than the red channel. Blue and green hold a lot of the contrast information in a picture, in fact green is so important that most cameras today have twice the amount of green pixels as there are red or blue ones. (Side note: Next time you do a black and white conversion of a portrait using a channel mixer set to monochrome, try reducing the blue and green channels and adding more red. This way you’ll get the skin much more smooth and reduce wrinkles. Your Significant Other might actually like that). What makes it more difficult is that the noise fingerprint is different for different cameras and for different ISO levels. So whatever noise reduction you want to do, the best ones are typically those that work on a per-channel basis (that’s often the case behind the scenes without you even knowing it) and first analyze the noise characteristics of each channel before applying noise reduction.

This noise analysis is not without issues though. Look at an image of a sandy beach and you’ll know that what you see on the ground is sand and not noise. Have a piece of software look at that same image and things might look very different. Telling noise from detail is not always easy, especially if you’re a computer program.

Obviously the by far best analysis can be achieved if you already know what the noise will look like and therefore can skip the analysis part altogether. And that’s what I believe the camera manufacturers do. They know what the noise characteristics of their camera’s individual channels are at what ISO settings. To get it even more precise, they might actually throw in a little thermometer and factor sensor temperature into the calculation. Ask your local astro photographer’s society, whose members might open up their cameras’ backs just to stick cooling Peltier Elements onto the sensors in order to minimize noise.

But I digress.

The camera manufacturers basically have the cards stacked in their favor. They know what others have to guess or find out the hard way.

Noise reduction itself is then done applying various linear or non-linear filters to the image employing gaussian or median functions or similar, or by applying them to individual channels or to the a and b channels of a Lab converted version of the image or … well, actually there are lots of different approaches to this and the specific approaches of different manufacturers are very often well-kept secrets. One general drawback of noise reduction is that the stronger it is, the more image detail gets lost. Think back to the sandy beach here. It’s very easy to turn the sand into a smooth plastic-like surface by overdoing the noise reduction. Another drawback is that the more sophisticated you want to get (e.g. the better you want to get the analysis) the more time the process will take.

Now how does that all tie into the differences between the camera manufacturer’s noise reduction vs. the Lightroom noise reduction. Remember, we are still looking for a way to mimic the camera manufacturer’s noise reduction in Lightroom as close as possible.

First let’s get one thing straight: he current noise reduction that both Canon and Nikon apply to high-ISO images out of their latest cameras, no matter if in-camera to the JPGs or if in DPP or Capture NX during development of the RAW files is clearly ahead of what Adobe Camera RAW (ACR, the underlying RAW engine of Lightroom and Photoshop) can do right now. But you can get close. Typically close enough for about 99% of my requirements. The trick lies in the right mixture of luminance noise reduction, chroma noise reduction and sharpening.

Luminance noise reduction is the noise that appears as individual pixels of the image randomly being slightly darker or brighter than expected. Think salt and pepper.

Chroma noise is noise where an individual color channel overshoots so far on individual pixels that the resulting image shows visible red, green and blue colored pixels scattered over the image. Think.. err.. psychedelic salt and pepper.. (sorry Robin, I had to use this at one point in time)

Both types of noise can be easiest spotted by a) zooming in to 100% (just hit Lightroom’s space bar) and by looking at a uniform flat grey surface in the picture. Think grey card.

Now let’s get practical and actually go through one image:


1. Set your camera to shoot a RAW & JPG image at the same time. This way you’ll get the unprocessed RAW file as well as the in-camera noise-processed JPG file for reference. Shoot the image at the desired ISO (I’ve done my first tests of this with the Canon 5D Mark II at ISO 3200, remember we’re still talking high-ISO here)

2. Import the RAW file into Lightroom

3. Open the JPG file in Photoshop (or in any other image editor that can give you an unaltered 100% view of the image) and resize both the Photoshop and the Lightroom windows so that you can see both at the same time.

4. It’s a good idea at this point to hit the little Y|Y button in the Lightroom toolbar below the image (if it’s not there, hit the T key first) to get a before-after view of your changes.

5. Go to the Lightroom Develop module and scroll down to the Details section in the right hand sidebar.

6. Now slide the Luminance slider to the right until the smooth surfaces of the processed image look similar to the smooth surfaces of the noise-reduced JPG that you have open next to it. It’s important that you look at the same part of your image in both Lightroom and the other app. Don’t worry if you lose a little edge detail in your picture, we’ll try to make up for that a little in the next steps.

7. Further up in the Detail section you’ll notice four sharpening sliders, Amount, Radius, Detail and Masking. These are our friends and they allow very fine control over the process.

sharpen.jpg(Side note on sharpening: sharpening does not add any detail to the picture, it will just make it seem more detailed. The software achieves that by raising the local contrast at visible edges in the picture. More sophisticated sharpening algorithms allow you to limit the sharpening effect in areas of the image where it might not be desired. Lightroom’s sharpening is one of those algorightms)

For the next steps it is crucial that your are zoomed in to 100% in Lightroom.

8. While holding down the ALT key, drag the Amount slider to the right. If you don’t see an effect right away, drag it all the way to the right to see what it does. Then go back to the left and drag it far enough to see a slight effect at edges in the image. Holding the ALT key while dragging the Amount slider will show you the luminance of the image only, making it easier to judge the amount of sharpening.

9. While holding down the ALT key, drag the Radius slider to the right. You’ll again see a different representation of your image, showing you how much of your image will be affected by the respective radius setting. I tend to end up somewhere between 0.6 and 1 here, and after getting used to it, I find the ALT key view very helpful.

10. While holding down the ALT key, drag the Detail slider to the right. This slider determines the amount of surface detail affected outside the edges. The more you crank this one up, the less smooth surfaces will look. I usually end up with this slider pretty far to the left.

masking2.jpg11. Last but not least, here’s the really important one. While holding down the ALT key, drag the Masking slider to the right and see magic happen. Okay, the black and white image you see now doesn’t actually look magic, but believe me, it is. The black areas represent areas that will not be affected by the sharpening at all, the white ones represent that areas that will. And that’s exactly in line with what I typically want my sharpening to look like: edges more than surfaces.

(Side note: very often you get away with only partially sharpening images, for example many portraits will work really well if you only slightly sharpen the eyes, lips, tips of the hair, rims of glasses and jewelery, but not the skin on the face. The viewer will perceive such an image as sharp, without seeing any over-sharpened pores on the face)

By using the Mastking slider in Lightroom’s sharpening section, you can very easily achieve that. I typically end up with the Masking slider pretty far over to the right.

If this is the first time you’re doing this, you might end up having to juggle some of those settings a bit and try different things to get it right, but believe me, it’s worth it. Just don’t overdo it with the sharpening. A little goes a long way.

And once you’ve found settings that are to your liking, simply save them as the new processing defaults for this ISO level and all your future images at the same ISO level will get the same treatment.

Is it perfect? No it isn’t. I will run super critical images through DPP. Is it good enough for most of my images? Yes. Absolutely. For now. Until someone manages to find a way to incorporate Canon’s secret noise reduction algorithms into Lightroom without making the workflow any more difficult than it has to be.

Author: Chris Marquardt

Chris Marquardt is an educator and podcaster. He wrote Wide-Angle Photography and is the co-author of The Film Photography Handbook and Absolut analog. He's the host of this podcast and a few others. Chris teaches photography all over the world. He is a regular on the TWiT Network.