Scanning Slides and Negatives: Comparing Film and Flatbed Scanners

"What kind of scanner should I buy to scan my photos?"

I answer this question fairly often, so created this comparison as solid data to refer to and for others' benefit. Here's the premise: You've got a large archive of photos and want to scan them--what's the best tool? First, if you have the option of scanning a print--4 × 6 or 8 × 10, machine or hand-printed--don't. Always opt for the negative or slide. A negative or slide can contain much, much more detail, contrast and resolution which gives a significantly better scan. A flatbed scanner with transparency adaptor is attractive because a lot of people already have them and/or they're much less expensive than a film scanner. How does the quality compare, though?

I used the Nikon CoolScan III, Nikon Super CoolScan 4000 ED (both film scanners), Umax PowerLook III and Epson Perfection 1670 (both flatbed) scanners and Hamrick Software VueScan to compare. The Fuji Velvia slide above was the test subject. This slide was chosen for its deep shadows, something most scanners have trouble with. All of these scanners can work with slides or negatives.

Technique

All scans are made at the highest optical resolution the scanner is capable of; I recommend always doing this. If you want a lower-resolution file, downsample the scanned image to fit your requirements. Do a little post-processing before the downsampling--levels adjustment and grain removal, for example--and you'll definitely get better results than just scanning at the desired resolution.

On all scanners this slide was scanned at 64-bit RGBI depth or 48-bit RGB depth, resulting in a 48-bit RGB TIFF file. (64-bit RGBI scans make use of a 16-bit infrared channel for Digital ICE or similar; it's only available on hardware that supports it.) Also known as 16-bit and high-bit depth scans, they hold a lot more information than a 24-bit (also known as 8-bit) scan. The extra information from a high-bit scan minimizes posterization when adjusting levels, for example.

These images displayed at 300% don't look too different, but the histogram easily shows how different they are. On the right is the 48-bit image, on the left is the 24-bit image. The smooth histogram of the 48-bit image equates to smooth tone. The jagged, spikey histogram of the 24-bit image means that some tones are missing from the image--particularly in the shadows. Don't throw away that information!

I used Hamrick Software VueScan 8.0.11 with all four scanners. I think it's substantially better software than what is included with the scanners because it's easy to use, very capable, and gives me complete control. It's also worth noting that its ease-of-use and capability makes it easy for a scanning novice to get a good scan; the same interface also allows an experienced scan operator to eek every bit of detail out of a scan. I used Adobe Photoshop to compare the scans.

Reviewing the crops below will only tell part of the story. The histogram is an important overview of the information found in each scan. If you're not familiar with a histogram, reading the brief Understanding Histograms at Luminous Landscape will cover the basics.

As the point of this comparison is to see what each scanner is capable of, the following results have received only minor post-processing: levels/curves adjustments to improve contrast (but not throw away any information), reducing the bit depth to a web-friendly 8/24-bit, and converting to the sRGB color space. Color hasn't been adjusted, the scan hasn't been resized, no sharpening has been applied. Raw and adjusted histograms are presented.

Don't Bother Scanning The Print

First, the disclaimer: if all you've got is the print, scan it. It's better than nothing. The point of this article isn't to tell you that you should never scan the print.

I had a 4- × 6-inch print of the test slide made at a local Wal-Mart, plopped it in the Epson Perfection 1670 scanner, and made a scan. The resultant file was 6,364 × 9,326 pixels and 339 MB large--that's 56 megapixels!

Comparing this scan to all of the others made with the slide, you'll immediately see how yellow/orange all of the leaves are. Additionally, the sky--the brightest area on the image--has mostly been blown-out to white. There is very little variation across the sky and other bright areas--+/- 1.5 RGB values (3 RGB values total variation). Compare to the blue sky of any of the below scans and it's obvious the print contained none of that information. Physically comparing to the print, the print does leave the leaves looking too orange and yellow, and the sky does look blown-out; the detail isn't there.

Shadow detail, on the other hand, looks pretty good, showing +/- 4 RGB values (8 RGB values total variation).

The end result is simply that there isn't much information in the print. You can't get a good scan out of it as compared to what you can get from the original slide.

Comparison

The idea behind the comparison is with the intent of archiving the photos on a hard drive, CD or DVD. With that in mind, capturing every bit of detail available is the utmost goal.

In the thumbnail at the right you can see three outlines. Those outlines represent the cropped area views that are available below at 100%. Since these scanners all have different capabilities, the 100% crop will vary in size.

Because of the very large size of these files, the original scans are not available to download. A small copy of the scan is available, giving an overall feeling to what the adjusted scan looks like.

Umax PowerLook III Flatbed Scanner

The Umax PowerLook III flatbed scanner with transparency adaptor is a 1200 × 2400 hardware dpi SCSI-connected professional flatbed scanner introduced in 2001. New its cost was about $700. It has since been replaced by a newer model.

This scan was made at 1200 dpi and 48-bit RGB. The resultant file is 1107 × 1646 pixels (roughly 1.8 megapixels) and 11 megabytes large. Printed at 300 dpi with no interpolation, this yields a 3.69 × 5.49-inch print.

The raw scan histogram (left) shows that the scan captured all (or nearly all) of the detail in the slide, though much of the information is in the shadowed area. The edited histogram (right) shows that there are a few spots in the darkest shadows where black already exists; midtones and highlights look good.

The below edited crops show well-defined thin branches and muddled leaves. Even on the original slide not much detail exists in the picnic table, though using the eye dropper tool at a 5-by-5 average shows +/- 1 RGB value (for a total variation of 2 RGB values) in the Adobe RGB color space--not enough to show any sort of detail. Colors are a little oversaturated.

My experience with this scanner is primarily limited to reflective scans, not transparencies. Reflective scans--prints, for example--are something this scanner is very capable of handling well, and colors are normally pretty neutral. Scanning the slide shows that this scanner has different characteristics with transparencies, however. Firstly obvious is that the scan is a bit oversaturated.

Epson Perfection 1670 Photo Flatbed Scanner

The Epson Perfection 1670 Photo flatbed scanner with transparency adaptor is a 1600 × 3200 hardware dpi USB-connected $100 consumer flatbed scanner introduced in 2003. c|net has a review of it.

This scan was made at 1600 dpi and 48-bit RGB. The resultant file is 1480 × 2187 pixels (roughly 3.2 megapixels) and 19 megabytes large. Printed at 300 dpi with no interpolation, this yields a 4.93 × 7.29-inch print.

The raw scan histogram (left) shows that the scan captured most of the detail in the slide, though much of the information is in the shadowed area and some of the darkest shadows were probably clipped. The edited histogram (right) shows that there are a few spots in the darkest shadows where black already exists; midtones and highlights look good.

The edited crops below show well-defined thin tree branches. Nothing is very sharp, especially obvious in the picnic table crop. Even on the original slide not much detail exists in the picnic table, though using the eye dropper tool at a 5-by-5 average shows +/- 1.5 RGB values (for a total variation of 3 RGB values) in the Adobe RGB color space--not enough to show any sort of detail.

My experience with this scanner is limited because I haven't had it for long. Most of my experience with it is reflective scans, which this scanner seems very capable of, though color is often slightly undersaturated. In this transparency scan, color is obviously very undersaturated, likely a result of a low-power diffusion light.

Nikon CoolScan III Film Scanner

The Nikon CoolScan III LS-30 is a 2700 dpi SCSI-connected prosumer film scanner from 1998. It has Digital ICE for removing scratches and marks on old film and is capable of multi-pass multi-scanning to pull more detail out of the shadows. Imaging Resource's brief review is still available. This scanner can often be found on eBay for about $100, making it a great deal--new it cost $700. It has since been replaced. by a newer model.

11/22/04 Update: It looks like this scanner is getting popular again because finding one on eBay for less than $250 is a tough chore!

This scan was made at 2700 dpi, 64-bit RGBI, and six passes were made. The resultant file is 3680 × 2505 pixels (roughly 9.2 megapixels) and 54 megabytes. Printed at 300 dpi with no interpolation, this yields an 8.35 × 12.27-inch print.

The raw scan histogram (left) shows that the scan captured all of the detail in the slide, though much of the information is in the shadowed area. The edited histogram (right) shows that the image details are well-spread through the midtones and that black and white are well-defined.

The below edited crops show excellent detail and sharpness, particularly obvious in the leaves and thin branches. Even on the original slide not much detail exists in the picnic table, though using the eye dropper tool at a 5-by-5 average shows +/- 4 RGB values (for a total variation of 8 RGB values) in the Adobe RGB color space--enough to make up some speckled wood shadow detail.

Extensive experience with this scanner tells me it's very capable of anything I throw at it. Printable shadow detail is captured, color is accurate, and the result is sharp. I've made numerous 11 × 14-inch prints of files from this scanner that have proven just as excellent as traditional darkroom techniques.

Nikon Super CoolScan 4000 ED Film Scanner

The Nikon Super CoolScan 4000 ED is a 4000 dpi Firewire-connected professional film scanner from 2001, bigger brother and a generation newer than the CoolScan III. It has Digital ICE for removing scratches and marks on old film, Digital GEM to reduce grain, and Digital ROC to restore the color of faded film. In addition to being a single-pass multi-scanner it has a substantially larger dynamic range to capture more shadow detail than the CoolScan III. Imaging Resource's brief review is still available. New its cost was $1400--a steal when you consider that its output is comparable to a $5000 drum scanner like the Imacon Flextight 343. They can be found used for about $600. It has since been replaced by a newer model.

This scan was made at 4000 dpi, 64-bit RGBI, and six passes were made. The resultant file is 3676 × 5389 pixels (roughly 19.8 megapixels) and 116 megabytes large. Printed at 300 dpi with no interpolation, this yields a 12.25 × 17.96-inch print.

The raw scan histogram (left) shows that the scan captured all of the detail in the slide, though much of the information is in the shadowed area. The edited histogram (right) shows that the image details are well-spread through the midtones and that black and white are well-defined. Some posterization is beginning in the shadows, even in this high-bit scan.

The below edited crops show excellent detail, obvious in the thin branches and leaves--just like the CoolScan III. Also obvious is some chromatic abberation present on the slide but not picked up by the other scanners. Lens flare is well defined and evident, not a blob like the results from other scanners. Close examination shows the film grain in the sky and shadowed areas, also not picked up by the other scanners. Picnic table detail consists of +/- 8 RGB values (for a total variation of 16 RGB values) using the eyedropper tool with 5-by-5 averaging. The Super CoolScan 4000 ED's superior capabilities are clearly evident.

As with the CoolScan III, extensive experience with this scanner tells me it's very capable of anything I throw at it. Color is accurate, and the result is sharp. Unprintable shadow detail is captured (unlike the CoolScan III), which can be manipulated into printable shadow detail, improving the result.

4000 dpi scans are both a blessing and curse: the higher resolution scan lets me crop and enlarge more, but it also captures grain in the film. Fuji Velvia (ISO 50) grain can be captured and found in the image pretty easily, but can be satisfactorily removed and doesn't show up when printing. However, higher ISO film--especially ISO 400 and higher--has such dominant grain that it is difficult to remove and is visible on prints. My experience is that the lower resoltuion of the CoolScan III (2700 dpi) isn't capable of capturing the grain but does create a high-enough resolution print for most uses.

Making a 4 × 6 Print

The scanning examples above are fine for archival use. In my opinion, it pretty conclusively shows that the flatbed scanner just isn't that good. But how about for making a regular 4 × 6 print--is the flatbed good enough?

These files have been resized, received levels adjustments and minor color correction, converted to 24-bit and were sharpened--ready to take to Wal-Mart and have a print made.

Umax PowerLook III Flatbed Scanner

Click the thumbnail at the right for a 4 × 6-inch, 289 dpi (1156 × 1734 pixels) 890 kB file. 289 dpi is enough to make a convincing 4 × 6 print, so I see no need to upsample the image--which could further degrade quality.

The edited scan looks pretty good: everything appears sharp and colorful, though dark shadow detail around the picnic table and some trees will print black.

Epson Perfection 1670 Photo Flatbed Scanner

Click the thumbnail at the right for a 4 × 6-inch, 300 dpi (1200 × 1800 pixels) 750 kB file.

The edited scan looks pretty good, similar to the PowerLook III scan: everything appears sharp and colorful, though dark shadow detail around the picnic table and some trees will print black. Overall the image is a little dark. Brightening the image began to cause posterization, so I chose to have a dark image with even tones over a bright image with garish color.

Nikon CoolScan III Film Scanner

Click the thumbnail at the right for a 4 × 6-inch, 300 dpi (1200 × 1800 pixels) 1.2 MB file. Downsampling from the original scan resulted in 4:1 pixel averaging (every 4 pixels from the original scan were averaged into 1 pixel for the 4 × 6 print) creating a much sharper image than the original scan.

As good as the flatbed scans look at 4 × 6, this edited scan is substantially better. First, the image is substantially sharper, though in a physical 4 × 6 print this will probably be less obvious. Color is more subtle and varied. Dark shadow detail is evident; most will probably blow out to black on the print, but a little texture will be evident.

Nikon Super CoolScan 4000 ED Film Scanner

Click the thumbnail at the right for a 4 × 6-inch, 300 dpi (1200 × 1800 pixels) 1.2 MB file. Downsampling from the original scan resulted in nearly 6:1 pixel averaging (every 6 pixels from the original scan were averaged into 1 pixel for the 4 × 6 print) creating a much sharper image than the original scan.

This edited scan is slightly improved over the CoolScan III results and substantially better than either of the flatbed scans. The image is substantially sharper, though in a physical 4 × 6 print this will probably be less obvious. Color is much more subtle and varied and a little more saturated. Dark shadow detail is evident; most will probably blow out to black on the print, but a little texture will be evident.

Conclusion

The archive comparison shows the extremes of what each scanner is (or isn't) capable of by creating a scan ready to be archived--obviously the flatbed scanners fall very far short of what the film scanners are capable of. Both of the tested flatbed scanners are very good at their primary job--scanning reflective media. For that purpose, I've been very happy with what each is capable of. They're just not that great with transparencies.

I could continue the "real world" examples with an 8 × 10-inch print comparison, but I hope it's obvious that the flatbed scanners wouldn't be up to the task. When creating a 4 × 6 print, however, the flatbed scanner appears to be pretty capable. And of course, if you only want a web-quality scan any scanner is probably capable of giving acceptable output.

Given the cheap used prices for the Nikon CoolScan III, I'd hope most opt for one of these over a similarly priced flatbed scanner. It's an excellent scanner, capable of quality results--especially when there aren't deep shadows.