Introduction
Three years into the 21st century and the digital camera market is expanding
at a fierce pace. This is a major shift from the previous decade of consumer
photography as a largely mature market. What technology trends are enabling
this growth? The fundamentals of visual perception, the building blocks of a
digital camera, and image reproduction are all means of understanding some of
the key enablers of digital camera adoption. This article examines all of these
elements of the digital revolution, as well as analysing the battle between
digital and 35 mm film photography. Trends and future directions are also considered.
Consumers do not historically adopt new cameras for novelty value. Even for
users familiar with the functions and foibles of a camera, climbing the learning
curve with a new unit is a task. Controls are generally in different locations.
Learning how to access advanced features is time consuming. A new camera purchase
is often motivated by the need for replacement, the desire to attain new capabilities
or upgrade quality, or a wish for greater ease of use.
Most consumers take pictures to mark family events, capture vacation images
for an album, or to share photos of children with family and friends. A smaller
number of enthusiasts and professionals take photographs for work and for pleasure,
and often with much higher frequency. Photographers are sometimes reluctant
to change equipment. Nonetheless, digital camera sales are soaring.
In the case of digital photography, the appeals are many. The ability to instantly
see your new photograph is compelling. The ability to electronically share images
simply is also attractive. These factors have been somewhat tarnished by lower
image quality for a similar price point relative to film, and by a significant
dependence on the personal computer to get a photograph printed. These limitations,
however, are rapidly disappearing.
Market Background
Digital camera sales are rising quickly as costs fall and image quality increases.
In just a few years, digital cameras have gone from a niche to a major consumer
category. Infotrends Research estimates that 17.7 million digital cameras shipped
in 2001, representing over 20 per cent of the low-end camera market. In 2002,
that number was expected to climb to 27.5 million. Among Internet-connected
US households, the estimate is that 60 per cent will have converted to digital
cameras by the end of 2002. It seems consumers are buying into the digital imaging
story.
The digital camera market today is segmented into three categories: the under
$500 consumer cameras, the $500 to $1000 "prosumer" cameras, and the
$2000 and up enthusiast and professional cameras. Consumer cameras typically
have a 2 megapixel (MP) to 4MP sensor array, a 1.8" color LCD display, and a
3X optical zoom lens.
The prosumer cameras have a 3MP to 5MP sensor array, and a 3X to 10X optical
zoom lens. Prosumer cameras have manual adjustments to exposure and aperture,
larger and higher-quality optics, and often a sturdier camera body. Some of
these cameras also support adapters for wide-angle or other lenses.
The enthusiast and professional high end of the market covers cameras which
replace the single-lens-reflex (SLR), have interchangeable lenses, a flash hot
shoe, faster electronics, and the overall versatility and quality of an SLR.
Sensor size is 6MP to 14MP, with the exception of the Foveon-based Sigma SD9
at 3MP.
Human Vision and Still Image Perception
Color Perception
Humans perceive color using cone-shaped cells at the back of the retina. Most
color perception is concentrated in the fovea, a small area about 3 mm across
where all but peripheral vision is sensed. The three types of color cells present
in most people roughly correspond to their sensitivity to red, green, and blue
colors, centered at wavelengths of 588 nanometers (nm), 531 nm, and 419 nm.
These are not narrow-band sensors, however. Red and green cone sensitivity extends
broadly from 400 to 750 nm. The numbers of receptors are not equivalent. The
average person has only one blue cone for every 20 green and every 40 red cones.
Thus image detail in the blue is not perceived as well as in the green and red.
The eye contains roughly seven million cones. As a result of this physiology,
image capture can be tuned to bias capture of green details above blue, and
most sensors use an array with twice as many green as red or blue sensor elements.
As the absorption spectrum of red and green cones is quite similar, the eye
must perform a difference operation between red and green signals to distinguish
between these colors. Perception of light intensity is therefore significantly
stronger than perception of color.
Color perception is also highly adaptive to both intensity and lighting color.
Candelas per square meter (Cd/m2) are units used to measure brightness. People
perceive color in illumination levels ranging from about 10 Cd/m2 (near twilight)
to light-colored objects illuminated with full sunlight, at about 10,000 Cd/m2.
Lighting conditions also vary widely, from greenish fluorescent to reddish incandescent
lights and tinted lenses. Within a few seconds or minutes of transitioning from
one to another light, our mind adapts to the new illumination and our perception
of color changes appropriately. When wearing heavily tinted yellow lenses on
a snowy day for instance, the snow initially appears yellow, but within a few
minutes is perceived as white again. Both of these light adaptation capabilities
are challenges for all photographic methods.
Resolution Perception
The ability to resolve detail is measured in lines per unit of radial angle.
This is equivalent to lines per unit of distance if the viewing distance is
fixed. At 50 to 60 lines per degree humans lose the ability to see individual
lines, and a gray single color is achieved. This is a result of the density
of cone cells in the fovea, generally accepted as 120 cells per degree in adult
humans. One needs two cells to perceive a line and the white space next to a
line. At a viewing distance of 30 cm (or about a foot), the average person can
distinguish line spacing of 0.1 mm, or about 250 lines per inch in black and
white. Color variations at uniform brightness are perceived with substantially
less resolution.
Interestingly, human contrast perception is stronger at five lines per degree
than at lower frequencies. In other words, people see stripes more strongly
when the stripes are 25 lines per inch at a foot then two lines per inch at
a foot. This appears to be an anatomically driven result of the way edge perception
and contrast is perceived by the eye.
Photography Fundamentals
Understanding basic photographic terms is an important step when examining
digital technology's place in photography.
Focal Length
The focal length of a lens reflects the effective field of view of the lens
relative to the sensor size. A 50 mm lens produces a field of view of 27x40
degrees across the 24x36 mm wide active area of 35 mm film. A lens twice as
long, or 100 mm, will produce an image with half the view width, 20 degrees,
for the same film.
As the sensor size changes, the field of view changes with it - so a sensor
area of 12x18 mm with a 25 mm lens produces the same field of view as a 50 mm
lens on a 35 mm camera.
A 50 mm lens would create the same size image as a pinhole camera with the
hole 50 mm above the sensor.
Zoom lenses have variable focal length and require movable optical elements
within the lens.
Aperture
All cameras capture light through some type of hole, or aperture. In many cameras,
the size of the aperture is adjustable from a small percentage of the lens opening
to a large percentage. The aperture is always located at a point in the lens
where light from all elements of the scene are effectively mixed, so the total
light admitted is reduced uniformly throughout the scene. The aperture size
is characterized by F-stop. The F-stop is defined as the focal length of the
lens divided by the aperture diameter. For example, an f/8 setting for an 80
mm lens means the aperture of the lens is 80/8 or 10 mm wide. A smaller aperture
reduces the light-gathering ability of the lens, and increases depth of field.
Often, lenses with wider maximum aperture, and correspondingly greater light-gathering
capacity, are larger and heavier than smaller aperture lenses. A wider maximum
aperture gathers more light and produces a better image in dim conditions. However,
as the aperture increases in size, the range of distance for which objects remain
in focus decreases. This effect is called "depth of field" and is maximized
by a small aperture. Thus a photographer wishing to keep focus on both flowers
close to the camera (foreground) and a mountain range far from the camera (background)
will select a small aperture.
Speed
The sensitivity of a sensor or a film to light is measured using an ISO standard,
or ASA number. This number corresponds to the number of seconds required to
expose the film to a certain level at a specific lightness level. ISO 100 speed
film requires twice the exposure time for the same level of exposure as an ISO
200 speed film. Slower speed (lower number) films are more suitable for brightly
exposed scenes or scenes without rapid motion. ISO 25, 64, and 100 are common.
These films produce less grainy, sharper images, and require a longer exposure
time than faster films. Intermediate speed films such as 200 and 400 are the
workhorses of consumer photography, and represent a good compromise between
speed and grain. Very fast films, such as 800 and 1600, are suitable for capturing
images with lots of motion or at low light levels, but produce substantially
grainier and less detailed images.
These speed ratings are carried over into digital photography as well. They
allow photographers to set expectations appropriately for exposure times and
aperture. Many digital cameras can perform at a range of ISO settings, for example
at 100, 200, and 400. The lowest number represents the optimal setting, while
faster settings are performed by effectively multiplying the signal from the
sensor at shorter exposures. This multiplication makes noise more apparent in
the image.
Exposure
The exposure is the amount of time the film or sensor is exposed to light. A short exposure time is more forgiving of camera or subject motion than a longer exposure time. Most handheld shots suffer when exposures longer than 1/50th of a second are used, and most cameras have shutter speeds which can be varied between 1/1000th (or shorter) and 2 seconds (or longer) exposure times.
Digital Technology
This section explores the technical components responsible for digital image quality.
Sensors
Image sensors in today's digital cameras use CCD, CMOS, or Foveon's X3 technology. A sensor's major characteristics include pixel count, cell size, overall size, bit depth, and noise level.
Charge-Coupled Device (CCD)
The traditional color mosaic CCD cell consists of a photodiode and a capacitor
for storing charge. A photon striking the photodiode area of the cell releases
an electron which is stored in the capacitor until it is 'read out' of the sensor
array. The number of electrons in the cell corresponds to the light level in
that cell. Each cell has a light filter over it which limits the sensitivity
of the cell to one narrow range of photon frequency. A mosaic of three (or sometimes
four) such filter colors is used to obtain in a full-color image. The CMOS sensor
uses the same type of mosaic as the CCD, but the silicon fabrication technology
is closer to memory chips, with a similar architecture. A photodiode area is
also employed, but rather than store the charge in the cell, a charge-to-voltage
converter and an amplifier is used at each cell. The resulting sensor data is
more easily and rapidly read out. Most arrays are read out at 10 or 12 bits
per pixel of data. Most cameras for professional applications output and store
12 bits per pixels. All consumer cameras and prosumer cameras store images at
8 bits per pixel after digitizing sensor data at 10 or 12 bits per pixel.
The noise level of a sensor appears in random spots in images typically taken
in low light or with a fast shutter speed, showing up as misplaced or random
bright pixels. Noise is an unavoidable product of thermal noise in the system
(this is a reason why telescope cameras are sometimes cooled with liquid nitrogen.)
A signal is produced by a sensor cell even when the shutter is closed. Sensors
usable with less light (higher ISO setting) generally manage noise better (have
lower dark current values.)
Complementary Metal-Oxide Semiconductor (CMOS)
The CMOS sensor appears to be steadily taking market share from the CCD as
larger arrays require more complex supporting electronics for CCDs. Kodak's
latest professional digital camera, the 14MP DCS Pro 14n, utilizes a CMOS sensor,
as does Canon's 11MP EOS-1Ds. In the prosumer range of 4-5MP, the CCD sensor
is still predominant. While CCD is the more mature technology, CMOS requires
lower power and scales better to large arrays.
An overlooked detail to this point - a color mosaic filter rated at 3MP is
not capturing each color at each pixel location. In fact, while contrast is
captured at an effective resolution of something under 3 MP, color resolution
is an effective 1MP or so. Sophisticated post-processing of image data helps
here, but a real loss of sharpness is common when texture includes rapid color
variation. (For this reason, the Foveon X3 sensor at 3MP has about the equivalent
contrast resolution of a 5MP mosaic filter, and about the equivalent color resolution
of a 9MP mosaic filter. )
Foveon's X3
Foveon's X3 technology is the upcoming dark horse in this area. Foveon uses
a stack of three photodetectors at each pixel location, effectively tripling
the color data per pixel gathered. This translates into a fundamental advantage
in image sharpness per pixel. The image quality of the 3MP X3 sensor is roughly
equivalent to a 5MP sensor using a mosaic filter. In other words, the X3 sensor
produces nine million data points at three million locations (one for each color
and at each location), while the 6 MP mosaic filter produces a total of six
million data points. Currently, Sigma is the only vendor selling a camera based
on this technology, but others are expected to join sometime in 2003.
Photons and Pixels
Before counting pixels, the physical scale of light capture should be considered. The visible spectrum ranges from 400nm to 800nm, or 0.4祄 to 0.8祄. Thus the photoactive area of a sensor cell must be considerably larger than 0.8nm to capture detail. In addition, the limits of optical focusing are driven by the wavelength of light, so a penalty in sharpness is paid for attempting to focus down below several wavelengths in scale.
For this reason, professional-grade camera sensors use relatively large cells of 7祄 to 14 祄 square. Foveon utilizes a 9 祄 cell spacing in its X7 sensor. Some of the prosumer cameras being released currently have a 5MP sensor with a cell size of less than 3 祄, using the 1/1.8" CCD form factor (actually 7.6mm x 5.1mm, or 8.9mm diagonal). At this spacing, light sensitivity and image sharpness begin to suffer. For this reason, sensor array size should grow or manufacturers should adopt the Foveon approach (combining multiple color sensors in a single cell) to usefully accommodate more image data. The bulk of consumer digital cameras at 3 MP use a 1/1.8" CCD array with cell size of approximately 3.5 祄. Prosumer cameras featuring 2/3" CCD arrays (actually 11mm diagonal) with cell sizes of 3.4 祄 and 5 MP of resolution are also viable and widely used.
The latest set of enthusiast/professional digital cameras from Nikon, Canon, Sigma, and Olympus use a relatively large CMOS or CCD array, up to film format size of 24 mm x 36 mm. Cells size is above 6 祄, and light gathering power and sharpness are excellent. These have a different set of issues, as photosensitive arrays are not as sensitive to light hitting a cell at a glancing angle as they are from directly above. This problem is not as serious in film. As a result, differences in recorded intensity occur which must be compensated for with electronics, and additional image noise at image edges is possible. The principal advantage of this very large sensor size is the ability to use existing lens systems from 35mm cameras.
In order to balance sensor size, cell size, and resolution, Kodak and Olympus
have recently agreed to a 4/3" standard for sensor size. (This standard of measurement
is an old carry-over from video tube days; actually this sensor measures about
0.7" x 0.5" in size, or 22 mm diagonal.) It is capable of maintaining a cell
size of 6.8 祄 at 5 MP or 4祄 at 12 MP, and may be the basis for a wide range
of cameras including prosumer and enthusiast (SLR style) cameras over the next
few years The sensor size enables lenses to be far smaller and lighter for a
given focal length than their 35mm camera counterparts, without compromising
image quality in most lighting conditions.
Optics
The ability of a camera to resolve detail is fundamentally gated by the sharpness,
light-gathering power, and aberrations of the lens. Thus pairing a very high-resolution
light sensor with a very small lens will often result in fuzzier pictures than
high-quality optics paired with a lower resolution sensor. One common measure
of sharpness uses a resolution chart with lines progressively closer together.
The point at which lines can no longer be separately resolved is the sharpness
measurement for the optics and sensor together. For example, the 6 MP Canon
EOS D60 professional digital camera can distinguish detail to nearly 1700 lines
per picture height as measured using a standard resolution chart and in well-lit
conditions. The Canon PowerShot S40, a 4 MP prosumer camera, has nearly the
same ability to distinguish detail, out to 1400 lines per picture height, indicating
a well-balanced lens and sensor. The much smaller Minolta Dimage X, with a 3
MP sensor and a compact folded optical path, resolves detail out to a maximum
of about 800 lines per inch in similar tests. In this case, optical sharpness
has been traded for extreme compactness.
Optical sharpness is only one metric of an optical system. The lens must deliver sharp images over the entire range of zoom, must focus cleanly and quickly, and must focus light from red through blue onto the same optical plane (to avoid color aberrations). Distortion of objects caused by the lens should be minimal. Light-gathering power (measured by F-stop) should be as high as possible to enable photography in low light. The range of shutter speed and aperture should be wide enough to accommodate rapid action and long night shots. These features vary little between film and digital cameras in their nature, although the scale factor of a lens is changed by the relatively small size of most digital camera image sensors. Small lenses must be manufactured to correspondingly tighter tolerance.
Most of the consumer and prosumer digital cameras sold have an optical zoom
lens capable of ranging from a wide-angle setting to a zoom setting over a 3X
range of focal length. Using 35mm photography equivalents, many of these lenses
span from 35 mm to 105 mm in equivalent focal length. Most lenses in this range
will lose significant light gathering ability as the focal length grows and
the field of view shrinks. For example, the compact Powershot S40 mentioned
above has a light gathering ability, indicated by F-stop, of 2.8 at the widest
setting to 8.0 at the longest telephoto setting. Thus telephoto scenes must
be well-lit to capture color and detail.
A new trend is the rising zoom range and light-gathering quality of zoom lenses
in prosumer cameras. The newly announced Panasonic Lumix DMC-FZ1 advertises
a 12X zoom with F2.8 throughout the zoom range. Equivalent to a 35mm-430mm lens,
this represents a new benchmark in digital (and indeed in any) camera. This
was achieved using a very small 1/3.2 2MP image sensor. Olympus introduced the
C-730 a 10x zoom lens with a larger 3.2MP image sensor and a bright lens with
F2.8 to F3.5 over the range of zoom possible. The HP 850 has an 8x zoom and
a 4.1MP sensor, with a Fuji lens delivering F2.8 to 3.1 over the zoom range.
This trend is compelling as consumers recognize the versatility and value represented
by high-quality optics. While the cameras represented are a bit larger than
pocket-size, they are significantly smaller than a film-based SLR camera.
Storage
Once an image is captured with a sensor, it is converted to one of several
image data formats for storage. The most commonly used storage format is JPEG,
which reduces the image quality slightly while greatly reducing the image size.
Typically, an image from a 3 MP sensor (9MB raw) will occupy 750KB to 2MB after
JPEG compression. To accommodate the storage of a number of pictures, digital
cameras use small, removable Flash or magnetic (disk) memory. Non-volatile (memory
retained while power off) Flash memory is available in many formats, the most
popular of which are CompactFlash and Sony's Memory Stick format, both of which
store up to 512MB of data. IBM manufactures a small hard disk, the Microdrive,
which can be inserted into a CompactFlash slot and have up to 1GB of capacity.
These and other formats such as Secure Media, xD card, and SmartMedia enable
the practical storage and transfer of images. A glut of different formats exist
and the market will likely converge on one or two eventually.
Image Processing
Digital cameras have to perform a significant amount of data processing to take the raw data from a mosaic image sensor, interpolate the pixels to form a full three-color image, and adjust the color balance properly, before producing a satisfactory image. That image is typically sharpened, and then compressed using JPEG to form the final stored image. In addition to these operations, image sharpening, edge enhancement, and red-eye correction may be run on the image. In short, a digital camera manages a significant image processing chain of operations.
The quality of algorithms used can have a strong impact on the perceived quality of an image. For example, ability to correct for a wide variety of illumination conditions is a strong factor in customer satisfaction. The color balance algorithm used can require significant compute power. In fact, there is considerable art in selecting and tuning the algorithms used here.
For this purpose most digital cameras have a powerful CPU and digital signal processing (DSP) core in them. Higher end cameras employ a separate DSP chip, while consumer cameras may integrate signal processing and general processing onto a single camera processor chip. The speed of this system determines, among other things, how quickly images can be taken in succession. As sensors have gotten larger, cameras have relied more and more on DSP capabilities.
Effect of Spatial Compression
All digital cameras have some ability to perform image compression to allow more images to be stored and quickly transferred. JPEG compression will typically begin to produce barely noticeable artifacts of noise and loss of sharpness when the compression ratio exceeds 6:1 or 8:1. A ratio of 12:1 is generally an acceptable compromise between space saving and visual quality. This reduces the 9 MB from a 3 MP sensor down to 700 KB on average. Depending on the level of detail present and the choices made in JPEG settings, a noticeable loss in detail can be caused by JPEG compression. While this is unlikely to cause noticeable losses when a 4"x 6"print is made, the loss of sharpness can often be seen in an 8"x10" print when JPEG compression at 12:1 is performed on a 3 MP image.
Effect of Dynamic Range Compression
Professional cameras are often used in an uncompressed mode which preserves
sensor data at 12 bits. These RAW mode images are very large, but allow a broader
range of post-processing and enlargement options. Because these files are often
quite large - about 16 MB for a raw Canon D60 image - these cameras are more
likely to require large storage media.
Recalling that color negative film has an effective dynamic range of 1:1000+, or 12-13 bits, it is unfortunate that a file format and compression algorithm supporting 12 bit sensor data is not commonly used. Instead, consumer-level cameras internally compress the dynamic range from 12 bits to 8 bits. This can wash out detail in bright areas and/or lose detail in deep shadow, relative to film.
Digital I/O Interface
Once images have been captured, processed, and stored, they must be moved to
a facility for viewing, manipulating and printing. Most cameras today are equipped
with a USB port, capable of moving image data at up to 12 Mbps. This is adequate
for all but the largest images. One major problem of USB is that peripherals
are strongly differentiated from computers. This makes a direct connection of
a camera to a printer difficult. A new standard, USB On the Go, has been created
to resolve this but is in the early stages of rollout. In the meantime, "sneaker-net"
is a popular answer - removing the digital data cards and popping them into
a printer slot or PC adapter as necessary.
Some cameras have infrared data transfer, although this seems to be disappearing.
Several new cameras have been announced using Bluetooth wireless image transfer,
and this looks promising for convenient wireless transport of smaller images
from camera to PC or peripheral. Sony's recently announced DSC-FX77 uses Bluetooth,
for example. Initial versions of Bluetooth had peak data transfer rates of just
over 400 Kbps. But even at the latest speeds of approximately 720 Kbps, a single
2MB image would take 23 seconds to transfer. This is too slow for a convenient
digital camera interface. The advent of Bluetooth 2.0, with speeds ranging from
4.8Mbps to 12 Mbps, will make this I/O interface much more usable.
Film Photography
Film Background
Color film photography is a technology rich in history and closely tied to
the art world. Eastman Kodak popularized color photography after the introduction
of Kodachrome slide film in 1935. Color print photography using 35mm film grew
rapidly in 1961 after the introduction of Kodacolor II print film.
The central feature of film photography is the photosensitive silver halide crystal, embedded in a gel and painted onto a flexible perforated strip. Silver halide crystals, sensitized to different colors of light and layered in gelatin, react to exposure to light to form a latent image. Upon development, exposed crystals are oxidized and turn dark. Thus lighter areas on the negative film represent areas of less exposure to light, so a negative image is produced. The last decades have seen great refinement in the sensitivity and quality of film emulsions produced. Current generations of film may be considered the stable end product of this development.
An interesting perspective: Film photography is by its nature a digital media.
A grain of silver halide is either 'on' or 'off.' A broad distribution of sensitivity
in these crystals results in a wide latitude of exposure. The film sensitivity
distribution is roughly logarithmic in nature, corresponding well to the way
people perceive light.
Film vs. Digital Resolution
Photographic films are chemical photosensors with performance defined largely
by their sensitivity and the size of the photosensitive crystals in them. Silver
halide crystals in a typical ISO 200 film have a grain size of about 1 micron.
They are dispersed randomly throughout the layers of emulsion in a film. Faster
films use larger grains and hence have more noticeable grain texture in the
resulting images. Here is where a direct comparison to digital capture starts
to break down. There is no thoroughly satisfying correspondence between number
of film grains and number of pixels in a digital camera. Instead, it's best
to look at the ability to distinguish detail and the level of noise in an image
under differing conditions. Film is rated with a resolving power, or number
of line pairs which can be seen. For a color film, a 50 line pairs per mm rating
for a normal contrast image is considered a good rating. That said, a film must
function together with a camera and lens. The net resolution achieved is the
relevant measurement standard. An outstanding analysis of system resolution
discusses the relative resolution metrics for film and digital cameras. Using
this analysis, a quality color film and extremely strong 35mm film camera system
yields about 47 line pairs per mm.
This is very similar - slightly less than the 54 line pairs per mm of resolving
power of the Canon EOS 60D, and one has a first metric for resolution of film
versus digital. This 6MP digital camera and a quality color film camera stand
at rough equivalence using this standard of comparison.
Another crucial metric of film is exposure latitude. How far from the ideal level of exposure can the photographer stray before the image is lost. Color print films have a great deal of latitude relative to digital cameras. The ability to capture details in bright and shadow area is quite extreme. A color negative retains data across an intensity range of 1 unit to over 1000 units of light exposure, or 3+ decades. Color transparencies have a more narrow range, (slides) capturing data across a range of 1:100+ units, or 2+ decades. The images stored in a consumer digital camera, at 8 bits, have only 2.5 decades of data. An 8-bit per pixel digital camera has a range of 1:256. In this area, color negative film surpasses digital photography.
Visible grain is yet another applicable metric.
How large can an image be printed before the grain structure resulting from the capture and print process is noticeable at a certain viewing distance.
Once again, digital and film images have different artifacts, so a single metric is not easily employed.
To illustrate, click on the thumb nail images below to see portions of a 6MP Nikon D100 image, a scanned color film image, and an 3MP Olympus D-550 image, each measuring 10% x 10% of the pixels from the original image.
The images shown were enlarged to 400x600 pixels so detail can be easily seen.
|
|
Digital color image from a Nikon D100 camera. Image courtesy of Phil Askey, Digital Photography Review.
(click to view entire image)
|
|
|
Professionally scanned color film image from an Olympus Stylus (point and shoot camera), captured with ISO 64 35mm slide film. Image by Andrew Mutz.
(click to view entire image)
|
|
|
Digital photograph taken with 3MP Olympus D-550, captured at highest quality and representing 1% of the image area. Image by Andrew Mutz.
(click to view entire image)
|
All images show a loss of sharpness from the enlargement, but the appearance of grain in the film image represents a perceived drop in quality compared to the superior D100 shot, and the blur in the Olympus digital image is a notch below the quality of the film image.
While generalizing is always dangerous and these images are meant to be illustrative,
a rule of thumb appears to be that a well-designed 3MP digital camera can produce
images equivalent to a point-and-shoot 35mm using a 200 to 400 speed film. A
more sophisticated 6MP camera is needed to meet or exceed the quality attainable
with an SLR shooting a moderately fine grained film such as an ISO 100 print
film.
An extremely fine grained film can capture images with more than the resolution of the current crop of 6 MP cameras, but in principal should be more than equaled by the newest 11 and 14 MP professional digital SLRs offered by Canon and Kodak respectively.
Optics
Film cameras, as with digital cameras are only as good as the optics used to capture the image. The same considerations of sharpness, light-gathering power, and distortion apply. The generally larger active area of film translates into a typically larger lens for equivalent focal length.
Color Film Formats
Color film is most often used in either 35mm (24x36mm active area) or APS (16.7 mm by 30.2 active area) format. The smaller APS format allows production of smaller cameras at the cost of active area and resolution. In practice, both formats serve well for the production of consumer color prints up to 8" x 10" in size using color film rated ISO 200 or slower. Medium and large film camera formats are primarily used in studio photography and offer more film area and hence resolution.
Digital Photograph Trends
Resolution: How Much is Enough?
Print Applications
Given the sensitivity to contrast noted previously, a picture with true resolution
of about 250 dots per inch (dpi) will have no visible artifacts when viewed
from a foot away. This is largely born out in examining enlarged images, and
in fact most images are well reproduced at 200 dpi provided. Since color artifacts
are seen less easily than brightness artifacts, a color photograph with soft
color detail rather than sharp brightness detail can be printed with highly
pleasing results from 150 dpi resolution. An image with sharp contrast, such
as a photo of a black and white document, benefits from higher resolution.
Translated into typical uses for the cameras, a 3MP consumer digital camera
can produce highly pleasing 8"x10" enlargements, while the newest
generation of professional 6MP cameras is suitable for enlargements of up to
16"x20". That is, 150 to 200 pixels per inch capture-to-print resolution
is generally enough to produce a sharp-looking color print. Thus a 1600x2000
pixel image is suitable for an 8"x10" print. A well designed camera
such as the Nikon D100 can stretch this to 150 pixels per inch, as little resolution
is lost in the optics and image processing steps.
Today's 6 MP digital cameras, such as the Canon D60 or the Nikon D100, can
produce extremely satisfactory print enlargements to 16"x20" for softer images,
effectively at 150 dpi, and any content should print well up to 11"x14", or
about 210 dpi.
It's important to note that each sensor pixel location in a digital camera
(except for the Foveon sensor) is only sensitive to one color band, and images
are constructed by effectively guessing the color at intermediate locations
in the sensor. A 6MP camera typically has three million green pixels, 1.5 million
blue pixels, and 1.5 million red pixels. From these six million data points,
an 18 million pixel image is constructed. This type of sensor is called a mosaic-filter
sensor.
For the vast majority of consumer photo applications, printed and viewed at
4"x6", a 3MP mosaic-filter-sensor camera is a practical compromise, with enough
resolution to leave room for cropping or the occasional 8"x10" enlargement.
For professional use, where images are used in full-page print applications or larger, a 6 MP or higher mosaic-based sensor is required.
Screen Applications
Many people are viewing images on 60 to 80 dpi electronic monitors. Images seen on an emissive monitor are typically viewed from longer distance than a paper print, and users are accustomed to less visual sharpness. For these reasons, images of under 1MP are often more than sufficient for display. As higher resolution LCD monitors become more common, the expectation of resolution will likely rise towards the demands of print media.
Bit Depth: The Next Step
From Shadow to Light
Film maintains a real advantage over digital cameras circa late-2002/early
2003 where dynamic range is concerned. Even as consumer digital cameras have
achieved parity and more with point-and-shoot film cameras shooting ISO 200
and faster films, they have limited ability to capture scenes which range from
deep shadow to bright highlight. The following image, for example, has a white
picket fence which looks pure white and devoid of feature, since the camera
has allocated the available intensity data to the other features in the photograph.
Capturing More Dynamic Range
In order to capture the range of brightness needed in scenes like this, a sensor with analog-to-digital conversion (ADC) at more than 12 bits would be optimal. In desktop scanners, 16 bit ADC is now commonplace, but this has yet to reach digital photography while the race to larger sensor resolution has been progressing. The logical next step for digital cameras as spatial resolution stabilizes at or above 6 MP will be the incorporation of greater dynamic range of at least 12 bits and ideally 14 bits or more. Capturing more than 12 bits of dynamic range requires another approach at the sensor level.
In January 2003, Fuji announced their fourth generation SuperCCD. The SR version
of their sensor pairs one low-sensitivity and one high-sensitivity sensor cell
per pixel, enabling the capture of much greater dynamic range. This deliberate
tradeoff of resolution for dynamic range is the beginning of what will be a
significant and valuable trend for photographers.
SMaL Technology, a Boston-based digital camera firm, is marketing a line of
relatively low-resolution (VGA and 1.3MP) cameras which have a wide dynamic
range as a core feature. Branded as Autobrite, this technology is described
as proprietary.
Storing More Dynamic Range
Storing and representing this data will require modifying or replacing the
JPEG file format as currently used. The JPEG 2000 file format, a highly capable
but as-yet unpopular standard, does support lossy and lossless compression of
12 bit and 14 bit data. It is one logical choice for carrying this data. While
the format has existed for a few years, manufacturers of cameras have not yet
embraced it. A more computationally expensive set of algorithms and lack of
support in image viewing software are frequently cited as impediments to adoption.
Another approach to storage is being utilized by Eastman Kodak in their newly
announced DCS Pro 14n, a 14 MP professional digital SLR. The ERI format uses
JPEG for 8 bit data and places the extra dynamic range data in another portion
of the data file usually used for header information. Thus the format is downwardly
compatible with JPEG. It's a well executed approach, proprietary to Kodak at
this time. Tools to use the format are still pending.
Once the data is available in a standard form, a wealth of creative tools or automatic operations can be applied to select the brightness of the desired areas of a photograph. For display on monitors, a reduction to 8 bits must still be done at some point.
Printing requires effectively the same sort of compression. In fact, the dynamic range of photo paper is roughly 100:1. The extra bit depth of capture gives the photographer flexibility in allocating the data captured by the camera to this more confined range.
Painless Printing: The Key to Adoption
To this point, the focus of this article has been imaging performance. The
use of a camera for most consumers is more about convenience than about ultimate
performance. The original Kodak creed of "you push the button, we do the rest",
is still the gauge for usability.
This is an area where digital cameras have generally lagged behind. Until USB
made device connectivity straightforward, even transferring images from camera
to a PC was awkward. Printing pictures involved buying and managing supplies
of costly photo paper and ink, and carefully composing the images on paper to
maximize usage. The latest operatings systems, such as Windows XP, deserve particular
credit for simplifying image upload and printing functionality. At an average
cost per 4"x6" photo of $0.50-$1.00, digital photography has thus far been less
convenient and more expensive than dropping a roll of film off at the local
store.
Two rapidly growing trends are dissolving this barrier. The first, the adoption
of digital photofinishing via the internet and local stores, is expected to
be largely available sometime in 2003. These digital photofinishing stations
have already been widely installed. Once availability is common, outside printing
of digital images will be at least on par with film.
The second trend easing printing will be the combination of direct-to-printer
connectivity with lower media prices. While many PC users are comfortable managing
their photo-processing, a large number of amateur photographers would prefer
generating prints directly from the camera to the printer. Hewlett-Packard has
released the Photosmart line of printers, incorporating media slots for CompactFlash
and other Flash memory media. This enables the printing of both contact sheets
and various size prints directly from media. Other technology trends in this
area will involve direct connectivity from the camera. The adoption of a USB
standard designed for inter-device connectivity, USB On The Go , will be an
important element of this trend.
The cost of producing digital prints is beginning to decline. Digital photo
paper, last year priced at $0.80 per sheet, has dropped under $0.50 to as little
as $0.10 per paper-size sheet for discount brands. The cost of ink is also showing
signs of dropping. While a full-page photo printed with an HP printer will consume
over $1.00 of ink, based on HP's ink consumption rating, the more frugal Canon
printers reduce that to about $0.50. Ink costs appear to be the next frontier
in price battles between printer companies as more consumers seek to adopt digital
photo printing and see their digital prints coming in well above the costs of
sending data to Shutterfly or Ofoto.
Conclusion
The consumer and professional photographic space is in the midst of a renaissance,
as digital photography raises the level of convenience and flexibility of photography
for consumers. The rate of adoption is swift, and today's technology provides
the necessary quality and functionality for continued conversion.
Film maintains superiority over digital images in all but professional digital cameras with regard to exposure latitude and dynamic range. The next competitive feature in digital cameras after resolution will be bit-depth.
More convenient and lower cost printing will serve to motivate the near-total conversion of photography from a chemical to a digital process within the next several years.
If you have any questions or comments regarding this article, please do not hesitate to e-mail comments@codesta.com!
Reference Material
| Digital Photography News and Reviews |
 |
| |
| Digital Photography Review |
| |
Digital photography and imaging news, reviews of the latest digital cameras and accessories |
| Steve's Digicams |
| |
Consumers' resource for digital camera information and news |
| The Imaging Resource |
| |
Comprehensive resource for digital camera information |
| The Future Image Report |
| |
Digital Photography Industry News and Trends |
| |
| Market Background |
|
| |
| InfoTrends Report on Camera Sales |
| |
InfoTrends Digital Camera Sales in 2001 |
| Forbes Magazine Article on Digital Photography |
| |
Traditional photogaphy unit sales dropping |
| InfoTrends Press Release on Photography |
| |
Digital Cameras are Displacing Film |
| |
| Human Vision and Still Image Perception |
|
| |
| Spectral Sensitivity of the Eye |
| |
Kenneth Koehler's "Spectral Sensitivity of the Eye" |
| NIST Base units definitions |
| |
Cd/m2 represents Candela per square meter, a unit of brightness defined by the brightness of a perfect radiator heated to the melting point of platinum. |
| WebVision: Visual Sensitivity |
| |
WebVision at the University of Utah: Visual Acuity Metrics |
| |
| Photography Fundamentals |
|
| |
| photo.net Learning Photography |
| |
photo.net Learn Photography - basics of lenses, film, light, cameras, technique |
| ANSI Document Store |
| |
ANSI Photography - Colour Reversal Camera Films - Determination of ISO Speed |
| Better Digital Online on Digital Noise and ISO Control |
| |
What does the ISO setting do on your camera do? |
| |
| Digital Photography Technology |
|
| |
| Eastman Kodak Technology Feature: CMOS vs CCD |
| |
CMOS vs CCD sensors and the future of digital imaging |
| Foveon X3 Technology |
| |
The world's first full-color image sensor |
| Digital Photography Review coverage of the Four Thirds System |
| |
Olympus and Kodak confirm 'Four Thirds System' |
| Kodak's 4/3 inch 5mp CCD announced |
| |
Kodak announced that it is working in partnership with Olympus Optical Co., Ltd., to develop image sensors for use in the new 4/3 System |
| Standard Test Chart List |
| |
ISO Resolution Test chart displayed |
| Imaging Resource Review Camera Reviews |
| |
Canon EOS D60 Review |
| Color Balance in Digital Cameras |
| |
Color Balancing in Digital Cameras |
| |
| Film Resolution vs. Digital |
|
| |
| Norman Koren on Image Sharpness |
| |
Understanding image sharpness Part 7: Digital Cameras vs Film |
| Eastman Kodak Company Color Film Data |
| |
Sensitometric curves for Color Negative Film |
| Photograph reproduced with permission of Digital Photography Review |
| |
Nikon D100 Review Samples Gallery |
| |
| Digital Photography Trends: Dynamic Range |
|
| |
| Fujifilm Company Press Center |
| |
Fuji Film Company announces fourth generation of SuperCCD Technology |
| JPEG 2000 Standards Group |
| |
JPEG 2000 Background and references |
| Eastman Kodak Company Press Release |
| |
Kodak Professional Debuts Extended Range Imaging Technology |
| Dissertation of Kresimir Matkovic |
| |
Display Devices and Contrast |
| |
| Digital Photogaphy Trends: Painless Printing |
|
| |
| Imaging Resource News Release |
| |
Fuji, IBM roll out kiosks, digital labs |
| Hewlett-Packard Company Product Announcement |
| |
hp photosmart 7550 photo printer |
| USB Consortium |
| |
USB On-The-Go |