A few simple tips to get the best image quality from budget film scanners. Scanning film can be challenging, high quality dedicated scanners are expensive, and even entry level flatbed for film are pretty pricey. My preferred method is to scan film with a macro lens – perhaps I’ll talk about this method at some point in the future – but today I was going to mention some tips for using the basic 35mm film scanner units which are available on ebay/amazon for between £30-£50. These units all seem to be the same or at least very similar: they are powered by USB, scan to SD cards, and claim a resolution of 5Mpx. Maybe a new generation will come out with a higher resolution sensor, but there are plenty of issues with these scanners and the sensor is not the worst of them.
A good 5Mpx scan could be sufficient for digitising photos, especially if you don’t currently shoot film and want to digitise family photos from a few decades ago, as it’s likely that the photos where shot on basic point-and-shoot cameras. If you want to squeeze every drop of image quality from your Leica M7 you probably know that a £30 scanner isn’t for you. To be honest, I don’t think these scanner are for anyone, but if you happen to have one you might be able to eke out a little more image quality with this method.
Dust is the enemy of film scanning. Your film should be as clean as you can make it before scanning. Compressed air, or a hand pumped air puffer, can help to clean up your film and so can water (ideally deionised), however you should make sure not to scan your film wet. If you process your own film you’ll know that you can wash it with tap water and even dish soap, however tap water can leave streaks in your film from the minerals in the water (but if you just dropped your wedding photos on the carpet and they are covered in fluff this is probably a reasonable proposition – and you can re-wash your film if needed).
However, with this scanner you don’t need to clean your film at all. This film scanner was sitting in my dad’s computer cupboard for several years before I tried it, and all that time is has been filling up with uncleanable dust. Not only does the design of the scanner make it very likely to collect dust, it also makes sure that all of the dust is visible in your scans. The scanner has a slot from left to right which accepts the film tray, this slot doesn’t have a dust cover, so is constantly collecting dust. Inside the film scanner is a sensor and lens (just like in a digital camera) and a light source and diffuser. When the film is inserted the light source shines through the film. However, as the diffuser for the light source (a sheet of milky plastic) is very close to the film, dust on this sheet will appear in your scan.
If you scan photos with a dSLR you want the light source to be far from your negative so that any dust on the light source is out of focus.
The images above were ones which I had processed a few years ago. I thought that I should redo the process with a few intermediate steps to prove to myself that I had remembered all of the steps.
First you need to scan each frame multiple times. I did this two ways, the first was better and the second was faster. The first way is to scan the frame, then wiggle the tray around a little and scan again (you should repeat several times until you have at least 5 scans of each frame). The second method is to scan the whole strip 5 times and just hope that you moved the film holder a little each time.
At the moment you’ll have several images where the dirt is in the same place but the picture moves, we can swap that up so that the picture is in the same place and the dirt moves.
I found all of the images from the same frame. As you can see the scan area is significantly smaller than the 35mm film frame – another issue with theses scanners. To complicate issues, there are three different kinds of artifact which we want to remove, marks which are fixed in the scanner frame, marks which move a little in the scanner frame, and marks which are on the film. The distinction of marks which move a little in the scanner frame is important as if these didn’t exist we might be able to perform background subtractions, or use the image with no film as a mask to show where the dirt is.
The registered images show just how much fluff and rubbish is inside the scanner. For a more accurate alignment it is helpful to increase the resolution size (by interpolation), I doubled the number of pixels in each direction, and downsampled the image at the end of this whole process.
Next is where the magic happens. How do we keep the bits which are the same, but remove the bits which move? The median filter! Just as a quick recap, you probably know the median as a centre point estimate, you might use it like the mean if you have outliers in your data. Well, that’s exactly what we have here. Hopefully, if we stack each image on top of one another and look at each pixel we’ll have 5 pixels which are very close in value, and one which is bright white (the dirt is opaque and the negative has the intensity reversed). Now, if we sort the pixels by intensity and the select the one in the centre there is a very good chance our selected value will be close to the true intensity value on the film without the artifact from the dirt.
This process isn’t perfect of course, one issue is that if each image is very dirty then there might be many regions where all of the stacked up pixels happen to be in dirty regions. However, the more images we include the better we can remove the dirt. Also, we will average out noise and JPEG artifacts. For removing noise and JPEG artifacts taking the mean would be slightly more effective, however, for dirty regions which are close to 100% white caused by the dirt we would be left with 6 copies which all have 16.7% brightness all over the image – which might be worse than just one copy of them. Especially if you had to clone-stamp them all out.
The median filter cleaned up the image a lot, but there is still a lot of mess in the frame. Some of this is dirt on the film, or small scratches.
After touching up the image with the clone-stamp and adjusting the brightness we could say that we’ve got the most out of the scanner (and the frame). Zooming in you can the significant noise in the sky, and the halos around the birds caused by the sharpening algorithm in the scanner.
This film cost £1 (it’s Agfa Vista 200) and if you can’t see the grain in £1 colour film then there is probably something pretty bad going on in the scanning.
So, what are your scanning tips? Do you have one of theses scanners? Do you think you’ll use this technique?