Scanners and Digital Cameras

A variety of hardware is available that captures image data in a form that your computer ” and particularly Photoshop ”can understand. These technologies transform visual information into digital data. They work by translating light striking the sensors in the device into digital values that you can store and manipulate on your computer. As a preview of the issues this chapter will explore, Table 14.1 summarizes the factors that determine the quality of a scanner or digital camera.

Table 14.1: Factors That Determine the Quality of an Image-Capture Device
FACTOR	MEANING
Optical resolution	The maximum number of pixels per linear inch the device can create from information it gathers from a reflective or transparent image. (See the upcoming section Flatbed Scanners for a warning about optical vs. effective resolution.)
Scanning sensor	Charge- coupled device (CCD) or complementary metal-oxide semiconductor (CMOS). Some of the latest digital cameras feature CMOS sensors that are of very high quality.
Interface or driver	The computer program that controls the device s behavior.
Scanning modes	RGB, CMYK, Grayscale, CIE Lab. Scanners always scan in red, green, and blue, but the resulting file can be converted to and saved in these modes.
Document dimensions	The maximum physical size of the image to be captured in one piece. In the case of a flatbed scanner, that would be limited to the size of the bed.
Dynamic range	How extensively the device can sense render density in an image. The higher the dynamic range, the more tonal information the scanner can discern and gather. Dynamic range is measured in Density units (0 “4.0 Dmax), a logarithmic scale of light intensity.
Speed	The length of time it takes the device to capture an image.
Bit depth	The amount of tonal information the device attributes to each pixel. This is a reflection of the number of steps the tonal scale is divided into for each color scanned. Most devices are 8-bit, enabling a grayscale image to be recorded in 256 steps, whereas a 16-bit device will record 65,792 tonal steps for each color it scans .
Passes	The number of times the device must look at the image. Some older scanners require three passes to collect the data ”one each for red, green, and blue. Single-pass scanners are usually faster and more precise. Some modern scanners offer multiple samples in which the scanner makes multiple scans of the same area and averages them (or takes the cleanest sample) to create a pixel value for the scanned area.

Scanners

Scanners convert reflective images (photographic prints, drawings), film negatives , or color transparencies into digital data. A scanner measures the color content and tonal variations and converts the data it collects into numerical values. Each value sampled is assigned a value for its red, green, and blue components or, if the image is black-and-white, a grayscale value. The values are transferred to Photoshop, where they are displayed as color (or monochrome) pixels on the computer display.

Flatbed Scanners

A flatbed scanner bounces light off reflective art and directs the bounced light through a system of lenses to a sensor that converts color and tonal variations into digital values. Flatbed scanners produce images composed of numerical values that are later translated into image pixels. Compare flatbed scanners by evaluating the number of pixels produced per inch ”the more pixels the scanner can produce, the more detail the image can have.

When choosing a scanner, be aware of the difference between its optical resolution and its interpolated resolution (sometimes called effective resolution). Optical resolution is the amount of data that the scanner can collect optically. Interpolated resolution uses software to increase the resolution by generating image pixels with a mathematical algorithm. This can produce the undesirable softening of an image. Dynamic range is also a critical factor. A flatbed scanner can currently be purchased for less than $100 to more than $3,000 for professional-quality units. Usually the difference in price is in the actual resolution of the scanner.

Film Scanners

Instead of reflecting light off of a reflective image, a film scanner passes light through the emulsion layers of a film negative or a color transparency. The quality of film scanners is typically better than flat-bed scanners because they are of higher resolution and greater dynamic range. A film scanner s dynamic range determines its ability to distinguish variations of highlight and shadow detail.

If you re going to purchase a film scanner, look for one with a high dynamic range (3.4 “3.8 Dmax) and a high optical (actual) resolution (3,000 “8,000 dpi). Good film scanners for 35mm films range in price from $1,000 to $3,000. Film scanners are also available for 2.25 ³ — 2.25 ³ and 4 ³ — 5 ³ film, but these are significantly more expensive, typically $7,500 or more.

Although film adapters are available for some flatbed scanners, these devices don t produce the quality of scan that a dedicated film scanner can produce.

Drum Scanners

Drum scanners are analog/digital devices that have been used in the graphic arts industry for decades. Reflective art or transparent film is mounted on a drum that spins while the scanner s sensors record a small, tightly-focused area of image information. Older drum scanners output to process color separations, converting the information directly into halftone-screened film. Newer drum scanners convert the scanned information to pixel data and store it as a file.

The greatest advantage of a drum scanner is the fine detail it allows, producing images that are free of the optical flare that is common on most CCD sensor scanners. Drum scanners can produce extraordinary color separations, but they are quickly being replaced by high-end flatbed and film scanners whose greater productivity makes them more cost-effective in production. Only a few manufacturers are still making drum scanners. Another factor in the sunset of the drum scanner is the emergence of professional digital cameras.

Digital Cameras

Every digital camera contains a two-dimensional array of sensors that convert light from an entire image into pixels, and then into digital values. Light enters the camera through a lens and is focused onto the sensor array. Two types of sensors dominate the digital camera market: charge-coupled devices (CCD) and the newer complementary metal-oxide semiconductor (CMOS) chips.

CMOS sensors are becoming more popular, and will likely take over the digital camera market in the long run because of their lower cost of manufacture. Early CMOS sensors suffered from excessive noise, but the latest chips are successfully challenging CCDs common in most digital cameras. The most sophisticated CMOS chips have the advantage of having a discrete sensor layer for each color, eliminating the need to extrapolate color values from a grid of dispersed red, green, and blue sensors on a single plane.

The size (dimensions in pixels) and quality (dynamic range) of the sensor determine the amount of data that a digital camera can collect and the tonal range that the sensor is able to record without overexposure . Consumer digital cameras have become so sophisticated and inexpensive that they are now outselling consumer film cameras.

The appeal of digital cameras is obvious ”immediacy and quality are compelling factors that make these cameras so successful. Amateurs and professionals alike enjoy the ability to determine immediately whether their image is successful (Is it properly exposed? Did the subject blink?). After the image is recorded, it is ready for print or for another destination without processing and printing. The appeal of digital cameras to professional photographers is also knowing that the quality of the image is now superior to that of scanned film. The latest professional digital cameras have at least two f-stops of shadow information more than film, resulting in a better image.

Enter the megapixel. Cameras are now touted as having a certain number of megapixels, a measure of the count ”in millions ”of image sensors on the array in the camera. A 1500 — 2000 pixel sensor is a 3-megapixel sensor, and it will yield an image of approximately 9 MB when uncompressed (most digital cameras compress their images by using the JPEG method in order to save space on the internal storage card).

Consumer-quality 3-megapixel digital cameras now cost less than $300. (There is even a disposable digital camera available!) Semiprofessional ( prosumer ) digital cameras are moderately high-resolution ”up to about 6 megapixels ”that have a built-in optical zoom lens (2x to 3x zoom range). These cameras cost between $500 and $1,500.

Professional single-lens reflex (SLR) 35mm-style digital cameras with interchangeable lenses can produce digital images that are comparable to any 35mm slide (usually better). These are significantly more expensive than prosumer cameras, ranging from $2,000 to more than $30,000. Digital camera backs that attach to medium-format and 4 ² —5 ² film cameras are also available. The return on investment for a professional studio photographer is tremendous; payback of the initial investment can take as few as eight months based on the savings of film and processing not purchased. Add to the cash savings the ability to know that the photographer got the shot, and these seemingly expensive cameras are a bargain for those who make their living with a camera.

Professional SLR digital cameras (such as the Canon 1D, the Fuji S2, and the Nikon D1X) all use interchangeable lenses. These lenses are compatible with their film counterparts but some have a different effective focal length when used on a digital camera because the image sensors in most of these cameras are smaller than the frame of 35mm film in a conventional camera. The Canon EOS 1D uses a full-size (24 — 36mm) sensor, whereas the Fuji and the Nikon cameras require a conversion factor of 1.5 (both use Nikkor lenses). For the current Nikon and Fuji offerings (and some of the Canon digital cameras), a 50mm lens for a film camera will produce a 75mm equivalent image on a digital camera, and a 105mm telephoto portrait lens will produce the equivalent of a 152mm lens image.

The real effect of these slightly smaller sensors is that true wide-angle photography requires a lens that is exceptionally wide. Nikon announced in early 2003 a 12 “24mm zoom lens to answer the demand for a genuinely wide lens for their popular D1X professional digital camera.

Note	Many photo processors will digitize your traditional photographs. For a few dollars per roll, they will digitize your film and either ship the resulting files to you on a CD, upload them to a website, or both.

Digitizer Boards and Software

These special-purpose computer cards transform analog video signals into digital image data. Image sources for this technology include television signals and the output from VCRs and traditional (nondigital) video cameras. Such sources produce a picture screen (or frame ) by sending out a stream of image information. Digitizer video boards, sometimes called frame grabbers, wait until an entire screen has been received and then display it. Select the Capture function for your board, and you can grab one screen at a time.

This method of image capture produces the noisiest and poorest-quality pictures. Most video systems depend on motion to enable your brain to fill the gaps between successive frames , providing the illusion of greater image quality. As I m sure you ve seen, when you hit the Pause button on your VCR, the actual images are quite poor.

It is also quite easy to capture still images from video by using video-editing software. The latest digital video products ”Adobe Premiere and Apple Final Cut Pro and Final Cut Express ”enable still images to be captured from digital sources. Final Cut Pro s Freeze Frame function will capture a single frame (actually two interlaced frame cycles) and allow the export of the image in a variety of still photo formats. The results are slightly better than those captured through digitizer boards. Application of the Photoshop Video ’ De- interlace filter will usually improve captured video frames significantly by removing the obvious horizontal scan lines that are common in video frames (see Figure 14.1).

Figure 14.1: The picture on the left shows a still image captured from digital video over a FireWire connection. The one on the right shows the same image after applying the De-interlace filter in Photoshop.

Challenges in Digitizing

No matter which technology you use to capture your images, you should bear in mind the kinds of problems you might encounter when capturing digital images. Some of these problems are unavoidable. With others, preemptive measures can ensure that you create the highest-quality image possible.

Some scanners, including Nikon s line of Super Coolscan film scanners, can automatically repair many flaws found in film, such as fingerprints , film base scratches, and many emulsion defects (though not emulsion scratches). These scanners can also feature grain improvements and restoration of faded color in films scanned. Combined, these features will repair old film, remove harmful defects, and restore color to an image that has long been forgotten.

Figure 14.2 presents one example of this kind of capability. A scan from a 35mm transparency shows serious dust and fingerprints; this image is clearly unacceptable. But the same transparency scanned with Digital ICE enabled on the Nikon Super Coolscan LS4000 scanner removes most of the dust and makes the image better. (Cleaning the film would help a lot, too.)

Figure 14.2: (left) A scan from a 35mm transparency shows many flaws. (right) The same transparency scanned with Digital ICE enabled removes most of the problems.

Figure 14.3 shows a photograph of a United Airlines DC3 airliner that was taken in 1941 on early Kodachrome film; over time it has faded to a nearly unintelligible condition. But applying Digital Grain Enhancement and Management (GEM) and Restoration Of Color (ROM) on the Super Coolscan improves the image immensely. Almost astonishing results are possible with these extended capabilities, now available on several quality film scanners.

Figure 14.3: (left) This photograph has faded to a nearly unintelligible condition. (right) Applying Digital GEM and ROC improves the image immensely.

What follows are some of the problems you ll face in digitizing your images and some of the techniques you can use to solve them:

Noise Random digital values ” noise ”can inadvertently be added to and distributed across a digital picture. Noise is primarily a result of digitizing technology. Video capture boards create the most noise, scanners the least. To help limit noise, shoot digital photos or video with as much ambient light as possible. Image noise appears most in the shadows, and is often more pronounced in the blue channel than the other two channels. (See Figures C14a “c in the color section.) You can correct noise to a certain extent by applying the Despeckle or Gaussian Blur filters to the blue channel (see the Removing Image Noise and Artifacts sidebar).

Artifacts Unintentional image elements, or artifacts , are sometimes produced by an imaging device or a compression scheme such as low-quality JPEG compression. Make sure to clean your scanner glass, and dust your scanner bed and lid ”and the original photo ” to help prevent artifacts in your scans. Also be sure originals are free of dust, fingerprints, and other surface marks if possible.

REMOVING IMAGE NOISE AND ARTIFACTS

An image s blue channel often contains the most noise and artifacts. The blue channel is the least critical to the human visual system ”but it is still necessary for producing full-color images. Instead of blurring the blue channel directly, a better technique is to duplicate the Background of a flattened image. On the new layer that is created, change the blending mode to Color (from the pull-out menu on the Layers palette) so that detail is not affected and then apply the Despeckle or Blur filter conservatively on this layer. Merge or flatten the two layers into one. Indirect filtering in this case leads to superior results, while remaining artifacts can be removed directly in their respective channels with the Blur tool, instead of being globally filtered.

Resolution The resolution of a digitizing device is measured by the pixel count per unit of measure it produces: the higher the pixel count, the better the potential image quality. Keep this in mind when purchasing a scanner or digital camera. In general, when capturing an image, use a higher resolution than you will ultimately need. You can always use Photoshop s Image Size command to reduce the resolution to a more manageable final size. Be cautious about increasing resolution or interpolating (up-sampling). When Photoshop adds pixels by interpolation, the image might lose its contrast and sharpness.

Note	Chapter 13, Sizing and Transforming Images, shows how to determine scan resolution.

Bit Depth Bit depth refers to the capability of your scanner or digital camera to capture tonal information. The more bits a capture device allocates to each pixel, the more shades of colors can be identified in the original and captured with the scan. The most common color depth, 8 bits per channel (or 24-bit RGB color), can produce 256 shades of red, green, or blue each for a total of 16,777,216 digital values or colors.

High bit images consisting of 36-bit color (three channels with 12 bits per pixel) and 48-bit color (three 16-bit channels) can produce billions of color combinations at the cost of much larger file sizes. In Photoshop CS, these high-bit files can enjoy more of the image adjustments than in the past, so it is possible now to open or acquire, save and manipulate high-bit images in Photoshop without converting to 8-bit data.

Note

High-bit files need to be duplicated and reduced to 8-bit depth for output to print. This process will average out the larger amount of possible channel values in higher-bit data into regular 8-bit data, which will theoretically be superior to a file that has been scanned in 24-bit image mode. Some photographers swear by this process. Chapter 3, The Nature of the Beast, discusses bit depth in more detail.

Moir Patterns in Scanned Images Moir patterns are optical anomalies produced when one pattern is superimposed over another. They are usually created by scanning already-printed halftone images or repeating patterns in an original (this occurs, for example, when scanning photos of patterned fabrics or rattan furniture) that produce interference with the scanner s matrix of image sensors, as seen in Figure 14.4. Editing moir patterns can be problematic because they can vary significantly from image to image, depending on the scan and halftone resolution and the screen angles. You can avoid them by scanning only continuous-tone images or by using the scanner s descreening function (if available), described in the next section, Scanning Images. You can sometimes reduce moir patterns by changing the angle of the art on the scanner bed. After the image has been scanned, rotate it back with the Crop or Rotate Canvas tools. This technique requires some experimentation. Also note that eliminating the moir pattern from the most pronounced channel (usually the blue) can reduce or eliminate moir patterns.

Figure 14.4: Scanned directly from a printed magazine page, this image contains heavy moir patterns caused by the interference of the original halftone screens and the scanner s sensor.