On Scanning Negatives

Thoughts and Techniques

Previously on scaletta.org, I posted a long review of a negative scanner gadget (dead link removed). After using that gadget for scanning a ton of slides and negatives, I’m afraid I just can’t recommend it any more. The results for color negatives were OK (and only just OK after applying extensive post-processing as detailed in that article), but as I used it for slides it was just unacceptable. In particular, parts of the slide would blow out but then change to a sickly yellow color. Even when it didn’t blow out, the scans were harsh and overly contrasty. I ended up selling the thing off.

So after scrapping the scanner, I still had hundreds of old slides I wanted to digitize. So what did I end up using instead? That’s what this article is going to cover.

The Gear

OK, so to cut to the chase, I ended up “scanning” the slides (and now, color negatives) using my digital camera, which is a Nikon D5600 DSLR. Besides the camera itself, you need a few more pieces of gear to make this work:

A light source. I use a tracing pad that also works as a light table for slides and negatives. In a pinch, you could also just use a tablet or phone set to a white background. However, I’d say the dedicated tracing pad will give better results. I’ve found it’s brighter than an iPad screen, and it doesn’t show individual pixels.
A way to hold the camera over the light source. I use a tripod that can point straight down at a table where I’ve set the light pad.
A blower to clean off the slides. Every bit of dust or fiber will show up in the scan!
A close-focusing lens. A macro lens is ideal; I use the Nikon AF-S Micro Nikkor 40 mm f/2.8 DX macro lens. It lets you fill the frame completely with the slide or negative, and it’s sharp from edge to edge. It’s really ideal for this application, and it wasn’t super expensive (used). If you don’t want to buy a new lens, kit zoom lenses that come with cameras will typically work just fine—you just have to crop the image down a bit.
For scanning negatives, you may want a film strip holder to prevent the film from curling.

The Process

OK, scanning slides with your camera is pretty straightforward! You put the slide on the light table, point the camera at the slide, and snap away. I’ve found there are a few settings that will give better results:

Use aperture priority mode (“A”), and set the aperture to f/8. Almost every lens will be super sharp at this setting, and when you’re scanning a relatively tiny 35mm slide, resolving the details matters.
Turn on exposure delay mode and use a short self-timer (e.g., 2 seconds). This will make sure that the camera is rock solid on the tripod before it takes the picture.
Set the “Picture Control” to Neutral or Flat. Slides already have a “picture control” built-in, so if you shoot in something like Standard mode, the results will have too much contrast and will generally look bad.
Choose a fixed white balance, like Daylight. This is the same deal—the slide film already has a white balance, so you don’t want the camera messing with it. My light table is a little cooler than daylight, so I like to set a “custom” white balance by measuring the light table with no slide on it.
Set the ISO to 100 and turn off auto-ISO. This will make sure you’re getting the best image quality from the camera. Since we’re using a tripod, we don’t have to worry about camera shake.
I let the camera handle the exposure automatically, which works fine for most slides. Sometimes the camera tries to overexpose things, so don’t be afraid to go to manual mode if the results look weird.

Sample Slides!

Here are some sample scanned slides! These have been cropped and hit with with the “automatic enhance” command in Photos.app, and that’s all the work it takes.

Here’s a fun one. Nice view from the balcony, right?

Well, that one was actually taken with the camera’s white balance set to “Auto.” Here’s the same slide with white balance set to “Daylight” (fixed at about 5500 K) (and after blowing off the dust properly!):

This matches what the slide looks like when viewing with your eyes. The scene was actually closer to sunset! Wild. Use a fixed white balance!

But what about negatives?

Hold on to your butts...

All of the above has been about scanning slides, which are color positive images. Those are about the easiest thing for the camera to scan, because the colors of the captured images should match what we see with our eyes. Black and white negatives aren’t too much harder—you just need to “invert” the image using software. Easy-peasy and you can get good results:

But what about color negatives? These are much trickier. You can’t just invert the image, because most negatives include an “orange mask” that separates the colors. A simple inverted color negative results in cyan color cast, which is no good. Here is an example, with the orange-masked original on the left and an inverted version on the right:

More than that, once you get into inverting the image and removing the orange, how do you know what colors are “right?” Should your scans match the prints? Is there some standard you should follow?

No, not really! Back in the “old” days, when you’d get your pictures developed at a mini-lab, the technician would manually adjust the colors and exposure of each print! Plus, each type of photo paper has its own color casts and dynamic range that are very different from film! What this means is that there’s no real “truth” for what a positive image from a color negative should look like, and we should just focus on getting to a result that we’re happy with.

OK, so let’s get real. Here’s my technique for scanning and processing color negatives, developed by employing way too much experimentation.

Trust the process

Set up your camera the same as if you’re scanning slides, above.
Use a custom white balance measured off of a blank frame of the particular film roll you’re scanning. This will cause the camera to transform the orange mask to gray, which effectively moves the values of the colors closer together. As a result, you can retain a lot more dynamic range in a processed JPEG file, and there’s no need to record in RAW.
While you’re focused on the same blank frame, increase the exposure until the orange mask (now gray) is brighter—almost white without clipping. On a color negative, that orange mask color is the brightest color that will ever carry information. Anything brighter would just be from the backlight through the sprocket holes or around the sides of the film, so we can safely let it clip. By moving the orange mask closer to white, we are sure to capture the most dynamic range available from the negative. Here’s the earlier image captured using settings we like:

OK, now we need to process the image and invert the negative. I use Pixelmator Pro, but almost any competent image editing software will work. First, as before, we need to invert the negative. I also add some sharpening at this point.

OK, this is starting to get recognizable. Changing the exposure has improved the dynamic range, and because of the white balance, the cyan mask has been removed in the shadows. We still need to fix the colors. For that, you can use the “curves” or “levels”tool to expand each color channel independently to cover the full tonal range. Many tutorials will tell you to manually adjust the high and low points for each color channel manually so that each channel barely clips. This will work, but it’s simply not feasible for scanning more than a handful of frames. Instead, have the software automatically stretch each channel independently. That’s what it’s good for! In Pixelmator Pro, in the Curves panel, select “Auto Color” from the drop down menu. Here’s what you get:

OK, that’s pretty good! You can get similar results in the GIMP by using “Auto Input Levels” in the Levels window, or in ImageMagick with the -channel RGB -normalize command. A good hint is to make sure that the operation is only working on data from the cropped negative image area, and not on any sprocket holes, backlight, or other surrounding areas.

As a finishing touch, I added a new color adjustment layer and performed an “ML Enhance” machine-learning general image enhancement. I also erased some lint from the image with the Repair tool. Here’s the final image:

It looks great, and all of the editing is fast and mostly automated! When I compare it to the accompanying 25-year-old prints (which are on Fuji photo paper), it has much more vibrant colors and way more tonal range, especially in the shadow detail. What the heck, here is the final scan (on the left) compared to a scan of the print (on the right).

Plus, that’s a cool shirt, bro.