The stream of technical data that is routinely dispensed in promotional literature and by salespersons in photographic stores may appear daunting. This article introduces the prospective buyer to some of the hardware and software terminology that manufacturers and retailers use to describe the specification and performance of imaging products.

 
Introduction
At the heart of all digital cameras (digital compact and digital SLR) is an image sensor that employs solid state electronic processes to record a scene. In nearly all mainstream cameras the image sensor is a CCD (Charge Coupled Device) or CMOS (Complementary Metal Oxide Semiconductor) two - dimensional focal plane array
that is typically silicon based, though more exotic materials are used.
Today, CMOS image sensors that have added functionality and benefit from large scale manufacture are achieving the performance levels of image sensors founded on the optimised CCD technology that has dominated the camera marketplace.
The camera lens focuses an image of the scene (primarily reflected light) onto an image sensor comprising a mosaic pattern of light - sensitive picture elements or pixels. The image sensor effectively segments the image of the scene and generates data for the pixel spatial location, called the pixel 'address' and the response.
The response is proportional to the image related electronic charge that is released from silicon (material) bonds and accumulates at the pixel 'address' for the duration
of the exposure, under the control of an electronic and/or mechanical shutter.
Electronic charge is accessed using registers (systems of interline or frame transfer are typical) and reduced to digital data that is passed to an electronic processor (image processing engine) and then saved in a compressed and/or uncompressed file format on a removable memory card.
To preview/review images an integral colour LCD viewing screen that can also display image data and control menus is provided. The description is greatly simplified, though digital and film cameras share many components and features, the digital camera offers much more. Once downloaded onto a computer, electronic images
may be edited and manipulated, emailed, exhibited online, exported to a database
or spreadsheet, ..., and integrated into multimedia applications.

Image Sensors
The scale of excellence most used by manufacturers and retailers to distinguish their digital cameras is the pixel count. The effective pixel count is the number of pixels that contribute to each image (some edge pixels are unused). At the time of this writing, the effective pixel count of low - end (point & shoot) cameras is typically
8 MegaPixels, of mid - range (creative compact, SLR style, entry level DSLR) cameras is typically 8 - 12 MegaPixels and of high - end (for the professional photographer market) cameras is typically 14+ MegaPixels. Based simply on the pixel density, 
point & shoot cameras surpass professional cameras, though image quality is not the same. Even though the pixel count continues to rise, understand that MegaPixel count is not the whole story (1 MegaPixel = 1 million pixels).
The effective pixel count may be specified differently, using an array notation,
for example, my Nikon D80 has 3872 × 2592 pixels, that means an image width of 3872 pixels and an image height of 2592 pixels. Most cameras support image capture at lower resolutions, for example, the 6 MegaPixel camera may shoot at 1.5 MegaPixel resolution (this is useful for Powerpoint presentations or web publishing) to save storage space on the memory card. The baseline criteria are not set in stone, 
camera specifications change as a result of demand driven technological innovation.  
To create photo - quality prints a resolution of 300 ppi (pixels per inch) is the rule, however, this is printer/paper dependent, 200+ ppi is often acceptable.
While the visual differences in doubling the pixel resolution are modest, doubling the pixel resolution does provide scope to recompose (crop and enlarge). Notice that the displayed image is dependent on the absolute pixel count, when displayed at 100%
a 1024 x 768 pixel image fills a monitor screen set to 1024 x 768 resolution.

The spatial resolution or sharpness of an image (sharpness is a subjective quality) is determined by the pixel size (typically 2 - 8 μm, rectangular or square) and the pixel count, 8, 12, 14+, ... MegaPixels, you can resize/downsample to capture an image at lower resolution. Of course, for the same pixel count and fill - factor
(ratio of pixel (photosensitive) area to pixel area), larger/smaller image sensors have larger/smaller pixels. For any image sensor format, decreasing the pixel size enables higher spatial sampling (higher resolution image capture). Naturally, there are
trade - offs, though smaller pixels can record more faithfully the fine spatial detail (the change in brightness with distance) in a scene, they are less sensitive to light, (compared to larger pixels, lower SNR at the same ISO). 
As the pixel size is decreased, the degrading effects of diffraction and photon noise (fundamental limits) on image quality become more pronounced. Furthermore, increasing the MegaPixel count necessitates a larger file size, that lengthens the upload/download time.

The aspect ratio is the ratio of the format width to the format height and determines the relative dimensions of an image (photographic print). These dimensions relate
to the image sensor active area (that includes the pixels that contribute to each image). A square has an aspect ratio of 1:1 or (1.00), the 35 mm Full Frame format has an aspect ratio of 36:24 or (1.50), most compact camera formats have an aspect ratio of 4:3 or (1.33). My Nikon D80 DX format camera produces an image with
23.6:15.8 or (1.50) aspect ratio. Of course, you can resize a digital image (generate an image with any aspect ratio) using image - editing software. Traditional computer monitors have a 4:3 or (1.33) aspect ratio. Widescreen computer monitors have different (1.60, 1.85, ...) aspect ratios. HDTV has a 16:9 or (1.78) aspect ratio.


Some of the image sensors used in consumer and prosumer digital cameras are listed. FLM is the focal length multiplier (see Depth of Field). Notice the relative shapes and sizes (approximate) of the image sensor formats. 
                               

Format	Aspect Ratio	Diagonal (mm)	FLM
1/3"	1.33	6	7.2
1/2.5"	1.33	7.2	6.0
1/2"	1.33	8	5.4
1/1.7"	1.33	9.5	4.6
2/3"	1.33	11	3.9
1"	1.33	16	2.7
4/3"	1.33	21.6	2.0
Canon APS - C	1.50	26.7	1.6
Nikon DX	1.50	28.4	1.5
Canon APS - H	1.50	34.5	1.3
Full Frame	1.50	43.3	1.0

The field of view, (FOV), or picture angle is the acceptance angle of light from the scene and is determined by the focal length of the lens and the dimensions of the field stop. In the absence of a physical stop, the active area of the image sensor (the linear dimensions) decides the extent of the scene that can be reproduced.        

Field of view may be decomposed into horizontal (width) and vertical (height) components. Using simple trigonometry, the field of view is given by
                            

where SensorDimension is the horizontal or the vertical dimension of the image sensor. At close focus distances, the denominator is replaced by 2FocalLength(M+1), where M is the magnification.

The image sensors used for digital cameras are different in size (1/3", 2/3", ...) and type, there is no format standard (unlike the standards for film).

Given a 20 mm lens, the corresponding fields of view (relative to the diagonal) for 1/3" and 2/3" image sensor formats are 17.1° and 30.8°.
To demonstrate the field of view change with focal length, consider a 28 - 300 mm telezoom mounted on a full frame DSLR body (image format, 36 x 24 mm).
At the 28 mm limit the horizontal FOV is 2tan^-1(36/56) = 65.5° and the vertical FOV is 2tan^-1(24/56) = 46.4°.
At the 300 mm limit the horizontal FOV is 2tan^-1(36/600) = 6.9° and the vertical FOV is 2tan^-1(24/600) = 4.6°.
 

• Short focal length lenses (focal length shorter than the image format diagonal) give a wide FOV that can capture the broad extent of a scene. An extended depth of field means that most foreground and background features are
  in - focus. Short focal length lenses appear to expand perspective and may distort near subjects. Most cameras are supplied with a 28 mm equivalent lens for wide angle capture. For sub full frame cameras, 12 and 15 mm ultra wide angle lenses are available.
   
• A standard lens has a focal length of approximately the image format diagonal and gives a FOV similar to the central FOV of the human eye. For the 35 mm format the focal length of a standard lens is 50 mm.
   
• Long focal length 'telephoto' lenses (focal length longer than the image format diagonal) give a narrow FOV that can isolate and (magnify) the scene detail.
  A restricted depth of field means that most foreground and background features are not in - focus. Long focal length lenses appear to compress perspective. Perspective is determined by viewpoint not focal length.
   

Camera Sensitivity
The ISO equivalent speed is a measure of the (camera) image sensor sensitivity to light. Like traditional film speed that is assigned a scale, ISO 25 (slow) to ISO 3200 (fast), at low ISO equivalent ratings the image sensor is less sensitive to light, at high ISO equivalent ratings the image sensor is more sensitive to light.
The ISO range 50 to 400 is typical for compact cameras (with smaller image sensors), the ISO range 200 to 1600 is typical for DSLR cameras (for some cameras the ISO range is extended both lower and higher). My Nikon D80 has an ISO range 100 to 1600 (high options, 2000, 2500 and 3200). The freedom to select an appropriate ISO for each image capture is a feature of digital cameras, however, noise (like grain in film) can become objectionable at high ISO ratings.

The noise performance of digital cameras is a subject much debated. Sources of noise may be classified as sensor noise or non - sensor noise. Sensor noise includes dark current (temperature dependent) noise, read noise associated with the signal processing electronics and spatial (pattern noise variants) noise due to pixel to pixel non - uniformities associated with the manufacturing process. The most significant non - sensor noise is photon (shot) noise that determines the fundamental limit on noise performance. Photon noise is present at all times and best understood as a natural variation in the arrival (and the subsequent detection) rate of photons (light) from the scene. This is a property of nature, not a property of the camera.
In digital images, the above sources of noise are discernable as luminance and colour variations (speckles). The significant noise contributions (three noise regimes) are illustrated using the generic photon transfer curve.
                                

The most fundamental measure of digital camera performance is the sensitivity or
signal to noise ratio (SNR) that is dependent on the total noise (the sum of all sources of noise). The sensitivity (SNR = 1) is limited by the total noise. For high quality reproduction, the spatial resolution and signal to noise ratio of digital images must be at least 'as good as' film images.

Photon noise and the full well capacity (FWC) limit the maximum SNR that can be achieved, the FWC is the maximum charge (measured in electrons) that each pixel can store before saturation and is related to the pixel size. For most quality cameras, compacts and DSLRs, the noise performance is limited by photon statistics (photon noise is proportional to the square root of the average number of photons). At pixel level, photons are converted to electrons, SNR_electrons = √η x SNR_photons, and
SNR_photons = √N, where the conversion factor η is the quantum efficiency
(η < 100%) and N is the average number of photons. On condition that noise is not visible, electronic gain to increase the ISO rating (like pushing film) may not significantly degrade the image quality. The maximum ISO rating (highest sensitivity) is determined by the noise performance and the lowest SNR that is practical. By careful design of circuitry and optimisation, one or two sources of noise dominate the noise performance.
Even so, compact cameras and DSLR cameras do not routinely offer the same image quality for the same pixel count. In general, compact cameras have smaller image sensors and smaller pixels. Typically, smaller pixels have smaller FWCs, have
lower SNRs and a reduced dynamic range. The term dynamic range (the ratio of the FWC to the noise floor, characterises the capability of a digital camera to record the darkest and lightest (black to white) regions of a scene.
At high light levels, photon noise is the limiting factor, at low light levels and for
long exposures, read noise and dark current noise are the limiting factors.
Typically, the SNR of high - end DSLR cameras (FWC ≈ 50,000+e @ ISO 100) is
more than 3x the SNR of low - end compact cameras (FWC ≈ 5,000+e @ ISO 100). High fill - factor, low cross - talk image sensor designs using novel pixel structures and microlenses can enhance the sensitivity. Adding the charge from neighbouring pixels, binning, can increase the signal to noise ratio, but lowers the resolution.
However, the strengths and weaknesses of digital cameras are only revealed by spatial frequency domain analysis of a range of parameters.
Today, some digital cameras achieve the imaging and performance standards
(similar noise and resolution figures) of film cameras (understand that image quality is subjective). Nonetheless, be aware of unrealistic performance claims (especially for compact cameras employing small format image sensors) that appear to challenge basic physical principles.

For low light photography, both indoors and outdoors, higher sensitivity guarantees faster shutter speeds (to freeze motion for action and sports photography) to reduce the loss of definition due to camera and subject movement. Furthermore, higher sensitivity may circumvent the need for flash illumination, so that images appear more natural. At higher ISO ratings, because fewer photons are collected for a correct exposure (and electronic gain introduces noise), the corresponding SNRs are lower. Look for cameras that provide a noise reduction setting at higher ISO ratings (long exposures in low light conditions), where (at large print sizes) digital noise can become unacceptable. To minimise luminance and colour noise, noise reduction software is available as a stand - alone package or compatible plug - in for the popular image - editing applications.
 

Colour
In most digital cameras the image sensor is fabricated using silicon microelectronics technology. The image sensor is responsive to the broadband scene luminance and cannot distinguish colour. The spectral response of silicon extends from about
0.35 - 1.10 μm (1 μm = 1/1000 mm), beyond the region of the electromagnetic spectrum that stimulates the human visual system (0.4 - 0.7 μm). 
 

Most cameras include a short pass filter (hot mirror) to remove the near - infrared (NIR) component of the scene luminance. Use a TV remote control to check the infrared response of your camera. On condition that you can access the long wavelength response (about 0.7 - 1.1 μm), infrared photography is practicable. Remember, this is reflected solar energy, not self - emitted thermal energy. To prepare your camera for infrared photography attach an IR filter (Hoya R72, Wratten 88A, ...) that is opaque to visible light. Use the camera self - timer and a robust tripod, long exposures are typical. An alternative approach is to modify the camera, the IR short pass filter is replaced by an IR long pass filter, and the autofocus is fine - tuned to compensate for the focal shift. Several companies can provide this specialist service. The NIR reflectance of familiar subjects is variable, foliage has a high reflectance, water has a low reflectance. To finish, create a monochrome image of contrasting tones (even colour images) using an image - editing software package. A complete treatment of infrared photography is given in a forthcoming article.

Any perceivable colour is a mixture of the primary colours Red, Green and Blue suitably combined. For the reproduction of colour most digital cameras use a Red, Green and Blue primary colour filter that is registered (in a repeating sequence on the face of the image sensor) with each pixel. Each pixel is responsive to only one colour. The Bayer colour filter is configured as a 2 x 2 array of one Red, one Blue and two Green filters (the human visual system is most sensitive to yellow - green light) to capture 25% of the red light, 25% of the blue light and 50% of the green light. The mosaiced image data from an 8 MegaPixel camera consists of 2 MegaPixels filtered red, 2 MegaPixels filtered blue and 4 MegaPixels filtered green.

Given that one primary colour is recorded at each pixel address, theoretically,
the resolution and sensitivity are reduced by factors of 4 and 3 (compared to the panchromatic (B&W) image), however, spectral interpolation (to recover the omitted colour channels, the demosaicing algorithm combines the RGB responses from neighbouring pixels) is used to reconstruct the full colour image. There are many demosaicing techniques using non - adaptive or adaptive, hardware - software based interpolation. Interpolation is a mathematical technique used to estimate unknown data values from known data values, within the range of the known data values.
The most common implementations of the demosaicing algorithm use bilinear or cubic B - spline interpolation. Of course, there are trade - offs between colour fidelity, sharpness and image defects. RAW capture allows you to post process (on your PC) using more sophisticated demosaicing algorithms that introduce less errors.
Alternative colour filter configurations are the CYGM (Cyan, Yellow, Green, Magenta), the RGBE (Red, Green, Blue, Emerald) and the RGBW (Red, Green, Blue, White).

In a discrete sampling process, erroneous data (aliased data) may appear unless the signal is band - limited, consistent with the Nyquist - Shannon theorem. For cameras that use a colour filter array, moiré patterns and colour artefacts are sometimes evident. Most digital cameras use an optical low - pass (anti - aliasing) filter (based on birefringent crystal polarisation or phase delay) placed on top of the image sensor to (slightly) blur the image and remove the high frequency content, thereby exchanging sharpness for reduced aliasing. Some cameras use an optical filter and image processing software. Most image - editing software packages provide an
anti - aliasing tool that can reduce the most distracting artefacts.
 
There are many optical techniques for colour separation, most consumer and professional (compact and DSLR) cameras employ some variant of the Bayer colour filter (lateral colour filter). The pioneering non optical capture technology developed by Foveon is based on solid state electronic processes. To produce full colour
non - interpolated imagery, the Foveon X3 CMOS image sensor captures three
colour channels (100% of the light) at each pixel address using Red, Green & Blue photosensitive layers embedded at the absorption depth of the primary colours, similar to a film emulsion (vertical colour filter). Furthermore, the demosaicing process and the separate anti - aliasing filter are unnecessary. The Sigma SD1 DSLR uses an APS - C X3 image sensor.

Most cameras convert the response to light (scene luminance) from analogue form to digital form, typically 8 - bit data that corresponds to a tonal range of 256 (2⁸) levels on a scale from 0 (Black) to 255 (White). For photographic quality full colour images the three colour channels are coded as 8 - bit data (bit (colour) depth of 8)
to produce composite 24 - bit colour, about 16.7 million hues, more than the human eye can distinguish. Some high - end cameras use 48 - bit or 36 - bit colour palettes, images are pre - processed and downsampled to 24 - bit data.

The image histogram is a graphical representation of the distribution of tonal values (number of pixels of a given brightness) that can be used to optimise the camera settings and for post production. Learn how to interpret the image histogram
(all DSLRs display a histogram, some compacts display a live histogram and some
high - end cameras display separate R G B histograms). For an image that has no major highlights (image highlights flash on the LCD screen) or shadows the distribution of tonal values is centred (mainly mid - tones). For a high - key image, the distribution is displaced to the right (toward lighter tones), for a low - key image, the distribution is displaced to the left (toward darker tones). All scenes are different, all image histograms are different, there is no 'ideal' image histogram.
Use exposure compensation, typically 1/3 stop increments, (−ve) to darken an image that is clipped on the right, (+ve) to lighten an image that is clipped on the left. Modern cameras can analyse the scene dynamic range (the difference between the lightest and darkest regions of an image) to optimise the distribution of tonal values. Of course, histogram manipulation is possible using image - editing software.