bp default image

Better Scanning Techniques

How to squeeze the utmost quality out of your film/scanner combination.

Of the fundamentally different methods of scanning, making a scan from a reflection original-such as a print-is far easier than scanning a transmissive original, such as a transparency or negative.

For scanning reflection originals, the dynamic range that has to be encompassed by the scanner is modest: less than six stops (i.e., the difference between darkest and lightest is six stops). For transmissive originals, however, the dynamic range can be wide indeed, running to 11 stops or greater.

This is important for two reasons. First, it is extremely difficult to arrange lighting and optics that can penetrate the highest densities seen in modern films. Second, the long dynamic range makes heavy demands on the computer coding required. As a result, very inexpensive flatbed scanners can make superb reflection scans, whereas to obtain really good transmission scans you need relatively costly hardware.

Squeezing the utmost quality out of your film/scanner combination is simple to effect, but it costs in increased scanning time. If you are fortunate enough to be throwing away an old computer, see if it can still run the scanner (and printer too). Keep it, and forget about tying up your main computer. It is here that Apple Macs come into their own: One inexpensive cable is all you need to network two Macs together.

Scanning strategies
Modern scanners produce surprisingly good results. But a number of strategies can help you obtain the best-quality from whatever equipment you use.

Scanning in 48 bit: A number of modern scanners will scan into 16 bits per channel, i.e., 48-bit space (some into 36-bit space or 12 bits per channel). This is necessary in order to code for high maximum-density readings. Some will produce 48-bit files, while others only output files in 24 bit. If you have a choice, retain the scan in 48 bit. This, despite the overheads in terms of longer scan and larger files, is well worth doing for color transparency materials. Ensure you carry out color and tonal corrections while still in 48 bit. Unfortunately, even superior software such as Equilibrium’s DeBabelizer does not work in 48 bit, but most of the basic corrections are available in Adobe Photoshop and Corel Photo-Paint 10. Binuscan’s PhotoRetouch works fully in 48 bit. Once the basic corrections are done, you can (and indeed should) turn the image into 24 bit.

Multipass scanning: For dense negatives or high-contrast, high-density transparencies such as Fuji Velvia, choose multiple scanning if available. If your scanner driver does not offer multiple scanning, all is not lost: See if Silverfast (LaserSoft Imaging) or VueScan (Hamrick Software) drivers support your scanner by checking their websites. The process takes longer, but the results are well worth the wait.

Dealing with moire: If you have problems when scanning printed material, the most effective remedy is usually to scan at a much higher resolution than you need, then downsize later. Another trick is to try placing the original at an angle for scanning and correct the rotation afterwards. Start with an angle of about 15 or 30 degrees. This goes against the advice to avoid rotation (see my later note), but here both the angle and the losses due to interpolation contribute to suppressing moire.

Software or scanner sharpening: One of the choices presented by scanners is either to allow it to perform sharpening, or to do the work yourself in Photoshop. Sharpening (preferably unsharp masking) in the device that is acquiring the image for you may take longer, but it may give you better results than Photoshop. Keep in mind that the scanner’s interpolated preview image is not a reliable indicator of the effects of unsharp masking. The best approach is to determine if you can see any difference. Try scanning an image without sharpening, then with different levels of sharpening performed by the scanner. But remember that scanner-sharpened images will show dust and dirt equally as clear.

Maintaining a dust-free zone
Taking precautions to avoid problems with dust saves a great deal of time and trouble further down the line. So, keep your negative files clean and dust-free before and after filing them with film. Avoid stacking negatives flat or on top of each other; their combined weight will squeeze dust particles on the film into the emulsion or backing to create permanent damage.

When scanning, clean your film and handle it with utmost care. The scanner itself should be kept meticulously clean; keep it covered with a plastic sheet when it’s not being used. The glass platen of a flatbed scanner should be blown free of dust and kept clean of smudges and dampness. Use microfibre cloths designed for cleaning lenses to keep the glass platen spotless.

A number of film scanners use infrared light to detect dust specks prior to a mask-based removal of them. But silver grains scatter infrared very well, so dust removal based on infrared cannot work with black-and-white film, apart from those developed in a C41 process (such as Ilford XP2 or Kodak T400CN). Some systems will not work with Kodak Kodachrome film.

Input resolution
Poor numeracy skills combined with a confusing subject make resolution a topic that will baffle digital photographers for many years to come. One reason for this is that the term "resolution" is now used in at least five distinct senses:

* Input resolution: in points per inch (ppi), measures the fineness of details scanned-the frequency or density of data sampling.

* Output resolution: in pixel per inch (also ppi), measures the size of pixels in the output-the spacing of output pixels.

* Device resolution: in dots per inch (dpi), measures the fineness of output from a device such as a printer, or the density of device-addressable points.

* Sensor resolution: in pixels or megapixels, measures the total effective number of pixels available on a photosensor.

* Lens resolution: in line pairs per millimeter (lpm), this measures the fineness of detail that can be distinguished to a given contrast.

The other reason for confusion is that we are not used to the virtual unit. That unit is the pixel, and it has no size until it is output.

The central notion, therefore, is that you think pixels and you count in pixels. As you get used to working with digital images, you get a feel for whether the number of pixels you have in the image is sufficient.

Broadly speaking, given images of average proportions, images smaller than about 150 pixels long cannot show anything but the broadest structure of a picture. For Web use, your images generally need not be longer than 720 pixels-which is also adequate for smaller prints (i.e., half the size of an average postcard). For fair-quality prints and output to A4 size, you will work in the 2000 to 2500 pixel range. For good quality work, the numbers rise to image lengths of 3000 to 4500 pixels and more.

Output scaling
Another source of confusion is the variety of approaches that manufacturers take in designing scanner-driver software. Some require that you set a scaling or enlargement factor, while others ask for the output size-and some allow you to set both. Worse, as you change one figure, others mysteriously change or else you are forbidden to enter certain numbers-usually for reasons that are never explained.

Part of the reason is that scanner drivers are helpfully written to prevent users from creating impossibly large files, and to discourage odd resolution settings that will involve a great deal of interpolation. You learn as you go.

It helps to understand the relation between the scaling factor or output size that you set. Suppose you scan a 1-in.-sq section of a transparency. If the scaling factor is, say, 4x or 400%, the output size will be 4-in. sq. A factor of 6.73 produces a square of 6.73 in. Alternatively, if you set an output size of 4-in. sq, the scaling factor will be 4x or 400%. Personally, I find it unhelpful to set scaling factors and find it easiest to set output size.Note another complication, however. These are the chain-link symbols, which indicate that a pair of settings are locked to change in step (i.e., to keep their proportions). If you have already set a crop on the pre-scanned image and have locked the crop, the scanner software will only accept settings that maintain the crop proportions.

Now, if you set an output resolution of, say, 200 ppi, then the input resolution must change as you vary the scaling factor or output size. The input resolution increases to provide you with enough pixels to meet the output resolution that you seek. And when that resolution hits a limit it cannot increase to match, you will find that you cannot enter the scaling factor-it is forbidden. Say, for example, you want a 4-in.-sq image output to 200 ppi, then you need 800 pixels. The original 1-in. image must be scanned at 800 ppi.

Scanning silver-gelatin negatives
The scanning of normal black-and-white negatives (silver-gelatin images) presents special problems. If you have scanned black-and-white negatives at high resolution-say at least 2700 ppi-you probably noticed that the results were grainer than you expected. Or at any rate, they were much grainier than a scan from comparable color material. And if you tried to scan at even higher resolution, the problem became worse.

Fundamentally, the difference is that the silver particles making up the image scatter light, whereas the dye-clouds of color images absorb and filter light. At the same time, the lighting in a film scanner is highly directional; it is hard lighting. The net result is similar to printing with a condenser enlarger: Darkroom workers know that printed results display sharper-edged grain and higher contrast than those from a diffused light-source enlarger. Now, add this to the high-resolution raster of a scan and you create interference or aliasing between the film grain and the regular array of pixels. The resulting scan is artifacted.

Imagine a sharply defined grain: If it lies wholly within a pixel, it is accurately recorded. But if the grain only covers one pixel and half of another, it will register on both pixels, fi ling both. Thus, it appears to be twice its actual size. The result of grain/raster interference is that many silver grains appear larger than they really are. Similar problems can be encountered with color film, but they are, in my experience, not as marked as with black-and-white film.

You can try to reduce grain/raster interference through simple light defocusing-the classic botch for low-pass filtering. Set the scanner driver to manual focus if possible, or fool it by raising the film with a clear piece of film. This stops very fine detail from getting through. While this seems to run counter to common sense, it may reduce graininess without harming broad image features.

Another way to reduce the problem is to use chromogenic black-and-white films such as Kodak CN400 or Ilford XP2. These films create the image using dye clouds rather than particulate silver.

Variation of grain with color channel
Suppose we scan our black-and-white negative in RGB mode. We obtain what appears to be a grayscale image. Now, inspect each of the red, blue, and green channels. We observe that there are slight differences in the appearance of the grain. This is due in part to a phenomenon called Rayleigh scattering. Some of the light incoming to a film is re-emitted by the silver particles-this is scattering-but the amount of scattering varies with the wavelength of light. It increases as wavelength decreases, i.e., as the light becomes more blue. This is what gives us blue skies (scattering from small air molecules) and gives sunsets their glow. Long wavelengths are scattered less than short ones by dust particles so the reds and yellows penetrate the atmosphere, whereas greens and blues cannot.

Now, the differences visible in the channels in our black-and-white scans are consistent with this theory. In general, you will find that grain in the red channel is less than in the blue and green channels.

But it means you can take your pick. If you wish for the finest grain, select the red channel and simply delete the others. In theory, if you need even finer grain, you could try scanning through a red filter (but the quality fanatic should be aware that CCDs are usually filtered against infrared). To obtain coarser grain, of course, the blue or green channels will deliver.

In Photoshop, you can invoke the Split Channels command in the Channels palette to create three separate files, each one containing a separation. Just save the one you want to keep and discard the others. But of course ensure you do this to a duplicate file.

Tom Ang has been writing about photography for more than 20 years, and his own photography-both digital and film-has been internationally exhibited. This information has been adapted from his recent title, Advanced Digital Photography, and is courtesy of AmPhoto Books, an imprint of Watson-Guptill Publications (www.watsonguptill.com).

Rotating Crooked Scans
You can straighten up a poorly aligned scan in some scanner drivers and software by using the Crop tool and rotating it so it lines up, accepting the crop. Another more accurate method, available in Photoshop, is to use the Measure tool: Call this up and drag along one edge of the image. Then, rotate the image under Image > Rotate > Arbitrary. You will find a figure for rotation already entered in the box. Ensure the rotation is in the right direction (clockwise or counter-clockwise) and hit OK. Your rectified image is delivered. Note that the image canvas increases to encompass the whole image; you will then need to crop off the excess.