Salvaging Photos

SALVAGING BAD PHOTOS AND VIDEO

PART THREE - Dealing with Noise in Under-exposed digital images

Here is yet another example of detail being lost in darkness, due to an auto-exposure camera.
This time, however, there is an additional problem we are concerned with. An area of particular concern in digital photography and video is noise. One place where noise is especially problematic is in black or dark areas.

While the previous techniques have allowed me to expose more of the detail hidden in the shadows, it has also exposed more noise. Let's have a closer look below.

Here is a close up detail. I have also lightened this image quite a bit to make the problem clear.
Obviously, there is a lot of dirty color noise in these shadow areas.

Fortunately, there is a solution.

I briefly mentioned the Lab color mode earlier. Here is an example of how this proves extremely useful. In digital photography and video, most of the image detail lies within
the values, not in the colors. Also, most digital noise seems to be in the
colors, not in the values. Therefore, if we separate the values, or
lightness, from the color info, we can work on the color data
apart from the values.
Below is one of the two color channels, lightened considerably for
demonstration, which contains a great deal of noise.

So, we now need to remove as much of this noise as we can. There are
several filters that can help here. Blur helps somewhat, but should not be
overused, as it will cause colors to bleed outside their intended areas.
I find that a combination of the Median filter, the Despeckle filter, and
the Dust & Scratches filter do a pretty good job of reducing noise in
these channels. Using the "fade" option in Photoshop, I can mix these
techniques smoothly.

Below is the same channel, after I have applied a variety of these
filters. Not flawless, but the noise is greatly reduced.
I repeat this process on the other color channel,
leaving the "L" channel untouched.

I then return to RGB Mode.

Here is the same close up, after applying the above process.

And here is the result, after returning to RGB.
Not a complete remedy, but I hope you will find it useful in cleaning up some borderline photos.

If you are working with video, you probably already know that there is no Lab mode available in After Effects. For this, you will need to rely on a third party plug-in. Fortunately, there is an excellent product that does exactly this. Steven Walker has several useful plug-ins for After Effects, and one in particular does exactly what we need. The Channel Viewer plug-in allows one to work in a variety of color spaces, including HLS, HSV, CMYK, YUV, and Lab. This will allow one to apply the same solutions I have outlined here. The Channel Viewer is available as part of the Walker Effects Basic Edition, which at the time of this writing, is available direct from Walker Effects for under $100.

BACK PART ONE PART TWO

Return to Tutorials

Here is yet another example of detail being lost in darkness, due to an auto-exposure camera. This time, however, there is an additional problem we are concerned with. An area of particular concern in digital photography and video is noise. One place where noise is especially problematic is in black or dark areas.
While the previous techniques have allowed me to expose more of the detail hidden in the shadows, it has also exposed more noise. Let's have a closer look below.

Here is a close up detail. I have also lightened this image quite a bit to make the problem clear. Obviously, there is a lot of dirty color noise in these shadow areas. Fortunately, there is a solution.
I briefly mentioned the Lab color mode earlier. Here is an example of how this proves extremely useful.	In digital photography and video, most of the image detail lies within the values, not in the colors. Also, most digital noise seems to be in the colors, not in the values. Therefore, if we separate the values, or lightness, from the color info, we can work on the color data apart from the values. Below is one of the two color channels, lightened considerably for demonstration, which contains a great deal of noise.
	So, we now need to remove as much of this noise as we can. There are several filters that can help here. Blur helps somewhat, but should not be overused, as it will cause colors to bleed outside their intended areas. I find that a combination of the Median filter, the Despeckle filter, and the Dust & Scratches filter do a pretty good job of reducing noise in these channels. Using the "fade" option in Photoshop, I can mix these techniques smoothly.
	Below is the same channel, after I have applied a variety of these filters. Not flawless, but the noise is greatly reduced. I repeat this process on the other color channel, leaving the "L" channel untouched. I then return to RGB Mode.
Here is the same close up, after applying the above process.