Noise, what is it? We can answer that quite simply: Noise is anything that shouldn’t be there. We could get quite technical here but why do that? Would it help you to combat it? Probably not. So let’s just keep the simple explanation of it.
Noise generally has two characteristics, Chrominance noise and Luminance noise. Chrominance noise looks like random color patterns, and luminance noise will look more like grain or grainy pattern.
Knowing what causes noise helps us to prevent it in the first place. (Note, noise is always present in digital camera it is just the amount of it that can cause a problem in your image)
Some of the things that cause noise are:
- ISO noise
- Long exposure noise
- Underexposure noise when brought up in brightness
- Compression artifacts from JPEG lossy images
We try to keep these at a minimum or eliminate them by:
- Using the lowest ISO possible for the scene and conditions
- Keep your exposure times short unless you are trying to show movement such as silkening the water in a waterfall with along exposure or trying to capture star trails
- Getting a correct exposure for your image and not raising the exposure in post, in fact it is adviseable (in a standard image, not HDR since there are multiple exposures in an HDR) to slightly overexpose the image (without blowout) and then reduce the exposure in post, therefore lowering the noise into the shadows.
- In an HDR Image you make sure you get the correct exposure both by correct metering of the scene for your exposure sequence. And also getting sufficient numbers of exposures for the scene so no one exposure needs to be raise in exposure significantly during the tone mapping process
- Painterly or Grunge style images can be more problematic when it comes to noise since they essentially make all tones close to mid-tones. Raising a shadow tone to a mid-tone will increase the noise of those pixels
- Use Tiff’s or PSD as your file type since they are either un-compressed or use lossless compression on the files. If using JPEGs using the highest compression number, therefore the lowest amount of compression.
For more about choosing the right number of exposures for your HDR, see the article here.
For more information on metering for low noise and better exposures, see the article here.
OK, so despite our best effort to combat noise, we still have a noise problem in our final HDR Image. There is still hope for your image with the use of some of the excellent noise reduction software that are available.
We’ll look at two of them, Neat Image and Nik Dfine 2.0
Neat Image 7.0
Neat Image can be had as a stand alone or a plug-in for Photoshop, It is available in 3 varieties, a free Demo version with limited capabilities, A Home Version and a Pro Version. The big difference between the Home and the Pro version is that the home version only works with 8bit images. The pro version can do, 8, 16 and 32 bit images. Therefore in the context of HDR, I would recommend the Pro version since we prefer to work in these higher bit depths.
(Note, for this test Neat Image 6.0 was used. A new version, 7.0 with some significant improvements is now available but I wasn’t able to get it in time for this test)
Upon opening your image in Neat Image, the panel has tabs along the top above your image preview the first one is for profiling the noise. Profiling the image will select an area with equal chrominance and luminance and absence of texture. This is both because that is where noise is most visible but also because we don’t want the texture in an image, say sand, to be mistaken for noise.
Once that profile is established you can save that profile which takes into account the camera model and ISO and use it for other images taken with the same camera and ISO later on or for batching images. Camera/ISO specific profiles are also available from the Neat Image website. This is a very helpful feature but is more useful with single image standard photography. With HDR we are introducing some abnormalities that may be beyond the standard camera profile and I prefer to do a profile for each specific HDR image.
After the image is profiled we can move to the next tab, the noise filter settings. We can choose the amount of noise reduction and then preview that at varying zooms. The base setting of 100% Chrominance Noise and 60% Luminance noise works quite well for m most images without severe problems. (Note 7.0 has more advanced filter controls)
You want to strike a fine balance between reducing noise and loosing detail which can happen with too much application of noise reduction. You can tend to get a plasticy look especially to skin especially with an over application of luminance noise reduction. You can also raise the filter settings for each channel if you find the need to further reduce noise. You can also do some sharpening at this stage to bring back some of the edge detail that can be lost to noise reduction. I prefer to do that separately as my final step of my file preparation
Here are two examples of how well Neat Image works which is indeed very well.
This first image is of just normal high ISO noise exacerbated by a little underexposure. The Image was shot at ISO 800 on a Canon 5D
As you can see in the after side, Neat Image totally eliminated any noise present in the sky while leaving behind good detail. This is a 100% crop and before any Sharpening. So some of the softness in the rocks can be recovered later but the filtered side rocks are not much softer than the before side
The next example was taken from our lesson the other day on Local Adjustments. That image had multiple things wrong with it. Poor initial exposure and bracketing and then subsequent exposure lightening that really raised the noise up out of the darker areas. It has serious luminance noise but worst yet really terrible chrominance noise you can see in the almost speckled color appearance of the blue of the lifeguard stand
This image really required cranking up the noise reduction, but it still did a great jo0b of making this a usable image. I don’t think I would have wanted to go this far if it was a portrait of a person because I think it would have lost too much texture in the skin. But then again we don’t shoot many HDR portraits and it indeed made this image highly printable.
Checking on the rails they still retained most of their edge sharpness and one of the real telling areas is lettering. It kept that intact without some of the smearing that can occur on letters of high contrast to the background.
I honestly didn’t think I would be too successful with the image looking at it before I applied the noise reduction but I was pleasantly surprised.
One thing I would suggest. Just like we apply sharpening to an image based on the print size or the screen size it will be displayed at. I think that noise reduction too should e applied that way. If you are making a small print you can apply less noise reduction because it won’t be as visible in that size print. You should also preview the image using View>Print Size instead of 100% because you may be correcting for something that never will be viewed at that magnification.
To learn more about or purchase Neat Image Click HERE
Nik Dfine 2.0
Nik Define 2.0 is a Plug-in for Photoshop, Lightroom and Aperture. While the basic operation may be very similar to Neat Image, it has some very special capabilities when it comes to the application of the noise reduction. I’ll adress those in a bit.
In Photoshop, Dfine would be opened under Filter> Dfine 2.0. In Lightroom/Aperture it would be opened with Edit with.
Upon opening the screen is laid out like all Nik software for familiarity.A Preview centered left and then the working controls on the right. You have your choice of three screen mode, single image that the effect can be toggled on or off with a preview button, or you can do split screens of before and after split either vertically or horizontally. You can very the screen preview background of light, medium or Dark gray and have Zooming ability from fit screen to 100 & 300%
In your settings you can set Nik Dfine 2.0 to automatically profile the image and apply noise reduction. Again this software is looking for an even tones area with no texture and you ca save profile for specific cameras and ISOs for later quick editing or batching. You can have the software automatically select the areas with different sample points across the image or you can use a manual method and choose the points your self. The automatic measurement worked great.
After you measure, you then reduce. The reduce section has two controls set to 100% for chrominance and Luminance. You can move these controls higher or lower depending on how much noise you want or need to remove.
So as you can see up to now the two programs are pretty similar. The real difference with Nik Dfine 2.0 is in the application of the noise reduction. And just like when we were talking about Global and Local adjustments to images in a pervious post. Nik Dfine 4.0 has the ability to apply Noise reduction locally in 3 ways. First off it can apply the noise reduction globally and then on close of the panel, will automatically add a new layer for the noise reduction along with a layer mask that you can brush the noise reduction onto specific areas that you want. (Of course you could do something similar with Neat image but you would have to duplicate the layer before hand and add your own mask. Nik Dfine 2.0 does this all automatically (If you want the mask to stick on the layer, make sure you press the brush button before closing the panel)
But beyond just a layer and mask, Nik Dfine 2.0 allows you to do very different and specialized local applications of noise reduction; the first one allows you to apply noise reduction just to specific colors within the image. Very useful in say a flower image where there are very specific colors you may want to work on and leave the others alone and retain more detail.
The second method of selective application is through the use of Control Points just like those in Nik HDR Efex Pro that I reviewed here. I think I should take a moment now, as I probably should have done in the Nik HDR Efex Pro Review and talk more specifically how these control point work.
You place a point in an area that you want to selectively work on, and then you can change the size of the control point sample area. That sample area expands in a circular marquee, so it gets confusing looking at it that it might apply those adjustments in a circular pattern. That’s not the case at all. The circular area is just the area that the tones are sampled.
When the filter is applied it is applied to tonal areas all around the image that are of similar tonal values, but just on a declining scale as you move away from the control point. So similar tonal values outside the circular area will get some adjustments, but just to a smaller and smaller degree. Confusing at first but one you use it and see how the effects are applied especially if you use the Luminance and Chrominance mask that are available in Nik Dfine 2.0 it become VERY clear.
So just how did Nik Dfine 2.0 do on our sample images? Actually the results were very similar between the two softwares. In that they both did outstanding jobs. So that part is pretty much a tie. Nik just wins on the selective application side. Whether that is important to you is your decision.
Here is our first image the one with high ISO noise in the sky area.
For the tower image, automatic didn’t really do the trick but a quick switch to manual measuring method and sampling the tower did the trick
For more information or to purchase Nik Dfine 2.0 as part of the Nik Collection by Google