IntroDigImg

Intro to Digital Imaging

THE DIGITAL CORNER

This month I thought I'd start an introduction to digital imaging for photographers. I'll divide this topic into three areas: Image acquisition, image manipulation, and image output, one area per month.

What I mean by "acquisition" is getting a still image into digital form so that it can be manipulated with a computer. The world is, in a sense, analog. So acquisition always involves some kind of analog-to-digital conversion (ADC). This conversion may take place immediately (when the picture is taken with a "digital camera") or later (when scanning the film or print). In either case the image is presented to an electronic device which responds to light with an electrical signal. In most cases this is a charge coupled device (CCD). It is a common misconception that CCDs are digital devices since they are so widely used in devices which output digital images. In fact, CCDs produce analog signals where the voltage is proportional to the amount of light which strikes the sensor. This voltage is then converted to a digital representation.

CCDs can be either 1- or 2-dimensional. That is, an array of sensors that is one unit high by X units long, or an array of X units by Y units. 1-dimensional arrays obviously cannot acquire much of an image in one exposure. They are used in "scanning" devices (including some types of digital cameras) where the sensor array is mechanically moved across the image, making an exposure of X points at each step of Y motion, thus forming a 2-dimensional image. This is fine for film and print scanners, but rather limits the applications for scanning cameras. In fact, a big problem for studio use is that special lights are typically needed to avoid the flicker that occurs with most types of electric lights. A 2-dimensional array exposes all points of the image at once, and can thus afford a longer exposure, which averages out the variations in illumination. So why not use only 2-dimensional arrays? Cost. The cost of a CCD chip is roughly proportional to the area of the active circuitry. A 4,000-element 1-dimensional array can be used to acquire a 4,000 by Y image while a 4,000-element 2-dimensional array can capture only a 64 x 64 pixel image. In practice, 2-dimensional arrays used in digital cameras are usually between 640 by 480 and 1024 by 768 elements; rather low [ouch: Today's digital cameras have as much as 3000 by 2000 pixels!]. 1-dimensional arrays are widely used in scanners, but also appear in digital cameras. Despite the lower cost of these arrays, such cameras are generally quite expensive because of the professional applications they target. Large 2-dimensional arrays have been marketed, but the cost is very high. For example, the 4K by 4K camera back made by Dicomed sells for $45,000 and Kodak's DCS460 with roughly 3K by 2K resolution sells for about $25,000. [The equivalent to the DCS460 today sells for less than $2000.]

Since CCD cost is proportional to physical size they are generally quite small, especially the 2-dimensional arrays. This makes them awkward to use in conventional cameras (which is desirable to allow use of existing lenses) because the image area is much smaller than typical film frames. Some digital cameras based on 35mm bodies employ a modified viewing screen marked for the smaller area. Shorter lenses must be used to achieve a given image size at the same distance. This is generally not the case with scanning cameras.

CCDs are, of course, imperfect image sensors. Each element of an array may have slightly different sensitivity. The cells also produce a small charge even in total darkness and this "offset" level also varies from one cell to the next. Another problem is that noise from the clock signal, which shifts the charge to the next cell, gets into the signal. The technology has improved tremendously to compensate for these weaknesses, but they are still apparent in certain situations. As mentioned in last months column on scanners, the most common problem is noise in shadow regions (highlights when scanning negatives). The best scanners are usually what is called a "drum scanner", which does not use a CCD array. Instead, a single sensor element (various types are used) is exposed with mechanical movement in both axes. The film or print is placed on a drum which rotates while the sensor is moved along the length of the drum. Since only one sensor is used it can be a very good one without having much affect on the cost, particularly since the precision mechanics of drum scanners make them very expensive, and, as you might imagine, they are not very quick either.

Digital cameras may transfer images to a computer by a variety of media, including SCSI, serial, or parallel communications ports, PCMCIA or other types of memory cards, or removable hard disks [today it is typically via high speed serial links such as USB or firewire]. Since cameras with 2-dimensional arrays have limited resolution or tremendous cost, and cameras with 1-dimensional arrays have limited application AND high cost, it is most practical today to use conventional film cameras and acquire the images through a scanner [not true anymore: In most respects it is far more practical to use a digital camera, unless you need very high resolution]. There are many service bureaus which will scan images for you and deliver the digital files on a CD, removable hard disk, or other media. Kodak's PhotoCD is available just about everywhere, including 1 hour mini-labs. Standard PhotoCDs give you images at up up to 3072 by 2048 resolution and typically cost about $1 per image. Only 35mm film is supported, however. The "Pro" format supports up to 4 by 5 film and offers much higher resolution, but typical cost is about $20 per image.

No portion of this web site may be copied or reproduced in any manner without express written consent.