Colin Dobbyne – OR Networks Limited
download the entire free booklet here.
Introduction
This technical reference booklet comes to you with our compliments. As a valued customer, we want you to get the most from the systems we have supplied and offer you help and advice in making the right choices for your application. This booklet should prove to be of great use as a constant reference source.
Designed to give a general overview of common imaging and presentation terms, techniques, problems and solutions; this booklet will give you the necessary background information needed to acquire the best images and display them to their maximum quality
It includes more than just equipment specifications and encompasses electronics, optics and computer technologies relevant to image acquisition, transmission and display. Inevitably some of the contents of this booklet may come across to the nontechnically minded as being too deep in its explanation. However, there are a variety of issues that have first to be understood before valid purchasing decisions can be made in relation to the imaging and presentation systems as a whole.
The booklet also offers an understanding of systems design. Any given system requires being no more under-engineered than it does over-engineered, though it may be that some items should sensibly be over specified at the outset with a view to future needs and upgrades.
The writing of this booklet is inspired mainly by the lengthy phone calls from our customers and potential clients who have picked our collective brains to benefit from our expertise. Now we want to bring that information within reach of a wider audience. Existing clients can maybe better exploit the features of their newly purchased equipment or at least understand the why and the how of the features. Potential clients can start to focus clearly on their real requirements and can also read between the lines of certain specifications that the manufacturers may claim.
Those readers who are setting out to purchase all or part of a system need to ask themselves certain questions in order to help themselves make a better assessment of what they need.
We have listed some of the more fundamental questions below.
Some questions
What is the nature of the image being recorded? Is the subject still or moving? Is it well lit or poorly lit? Is my camera sensitive enough?
What lens requirements are there and how will this affect quality?
Is it a live video application or are the images to be recorded or digitized? Is it for data storage, electronic analysis, printing to hard copy or simply for display? Is it required in as much detail as possible or simply for recognition or draft purposes?
What existing equipment, if any, do you have and which, if any, do you propose to carry on using? What software programmes do you have or plan to use with your equipment?
Who is driving the system, yourself, others, many others? How foolproof or simple must it be i.e. relative to the lowest common operator?
Critical factors
What are the critical factors, if any? Is there a minimum resolution you are trying to achieve? Is the image for analysis, if so what tolerance can you accept?
What will you want from your system in future years and will your critical factors change? Will you wish to migrate to higher definition?
If there is a critical factor to be identified then this should be the starting point of designing the system, e.g. DICOM compliance? 1080p mimimum?
Specifications and Units
To enable selection of the products most suitable for your application, it is important to understand exactly what the technical specifications mean. To do this properly a basic knowledge of the defining methods and measurement units are essential.
Units
Unit | Symbol | Definition |
lux | Lx | The derived SI Unit of illuminance; one lumen per square metre.
The amount of light falling on a square metre per second. |
NIT | Cd/m2 | The measure of brightness, typically of an LCD screen. Candela per sq metre is the SI standard measurement, commonly called a nit. |
decibel | dB | A unit for comparing two values. For example, two voltages |
Pixel | One light-sensitive element of a CCD (charge coupled device) array as used in cameras. Also referred to as the smallest unit in a digital display. | |
f-number | F | The ratio of focal length to lens aperture.
The larger the f-number the smaller the aperture and so the less light is passed through the lens. |
Volt | V | The derived SI unit of voltage. |
ANSI Lumen | The preferred measure of “brightness” for a projector as opposed to lux, it refers to the amount of light falling on a defined area, more representative of a projection screen. All projectors use this measure and so it offers a convenient comparator. | |
Bits, kilobits, megabits | b, kb, mb | The bit (b) is the fundamental on/off state of a single binary unit variable that makes up all digital data values. A bit has a value of 1 (on) or zero (off). A kb, is a thousand bits, a mb, a million. |
Bytes | B, kB, mB | Bits are conveniently grouped into a Byte, 8 bits. The basic value for digital information. The value range is from 0 to 255. It follows then, that a kB, is a 1,000 bytes, or 8,000 bits. |
Frequency | Hz | The SI unit for cycles per second. In video this often refers to how many full images per second are displayed. 50 Hertz equates to 50 fields per second. |
Specifications
Camera Sensitivity
This specification is sometimes quoted as minimum illumination. Below is a description of two methods, however, the minimum illumination figure is less useful for imaging applications as the picture quality is often quite poor.
Method 1 (f number - a precise bench mark)
To quote the light level that can produce a full video signal at a given f-number,
e.g. f5.6 at 2000 lux.
This simply means that the camera will output a full normal video signal (100%) under a 2000 lux environment without any amplification or modification.
Therefore, the larger the f-number at a given brightness the more sensitive the camera.
Method 2 (minimum illumination - imprecise with variable parameters)
To quote the lowest light density that will give a “useable” signal output,
e.g. 1.5 lux.
This figure is with AGC on (Automatic Gain Control), the lens aperture wide open and often not even a full signal. AGC can be up to +18dB (8 times) or +20dB (10 times), the problem is that noise is amplified as well, see Video Signal to Noise.
Conversion
It is possible to convert from one form to the other using the following formula:
minimum illumination x gain = quoted illumination
(minimum f-number)2 (quoted f-number)2
What is meant by TV lines?
TV lines are the number of vertical black and white lines resolvable in the central region of a perfect monitor, that is a monitor that has no resolution limitation. The lines are vertical and are counted in the centre of the image only to a distance equal to the vertical height.
To convert from TVL to pixels
In graphics and imaging, and digital video systems, it is more common to think of resolution in pixels. An approximation to the horizontal pixel count is given by multiplying the number of TVL by 1.5 (on 4:3 ratio screens). However, this is the extinction point, so in reality a graphics image with the same pixels width as the camera’s TVL is a good match.
This is measured in decibels (dB). Decibels are a convenient way of expressing a ratio between two values as it helps avoid very large numbers. In this instance, the values of the signal voltage (the wanted part) to the electrical noise (unwanted part).
Digital Signal Processing Cameras
Many cameras now process the video signal in digital form. The advantage is the same as with any digital signal ― there should be no degradation of signal. However, the number of bits available reflects the maximum signal to noise ratio and these are as follows:
DSP Bits | levels | Equivalent dB |
| 8 bits | 256 | 48 dB |
| 9 bits | 512 | 54 dB |
| 10 bits | 1024 | 60 dB |
| 12 bits | 4096 | 72 dB |
For the more observant, you should see a pattern emerging above.
In binary, for every bit in a string of bits, you double the number of possibilities.
So for every additional bit, you may add another 6dB to your maximum S/N ratio.
If your camera is 24 bit then your effective S/N is 24 x 6 = 144 dB.
Image Sensors
Currently, all CCDs (Charge Coupled Devices) are made from arrays of photo-sensitive semiconductive diodes called pixels (from picture element). These capacitive cells, or diodes, store charge in proportion to the amount of light falling on them. The charge is then transferred to a shift register (an integrated circuit for temporarily storing values) and the values read out, line by line, forming the eventual image. This process is repeated 50 times a second (field rate) in UK and Europe and 60 times a second in the USA.
To understand why CCD cameras work in this strange way, it is important to understand how the original cameras and monitors worked.
Due to the electro-mechanical limitations at the time of the first tube cameras and early monitors, it was necessary to build up a full image (one frame) on to a CRT (cathode ray tube) by having two scans (two fields). This gave rise to the process of interlacing odd and even fields. Each TV tube has 625 lines making up the frame, that is two fields of 312.5 lines.
CCD Types
The two main types of CCD array are interline transfer and frame transfer. Due to manufacturing and memory costs, interline transfer arrays are now more common place. The major difference is that for interline transfer the CCD charges for a field at a time and for frame transfer a frame at a time. This is significant for certain applications, particularly under stroboscopic illumination.
Single CCD Cameras
A camera using a single CCD can be colour or monochrome. If monochrome, then the array is normally only filtered to remove unwanted IR light, since the devices are naturally sensitive to IR. To produce a colour image, the CCD is filtered with a coloured pattern, the simplest of these is an RGB striped filter, see below.
There are alternatives to the striped filter,
for example mosaic patterns with complementary colours instead of primary. However, whichever filter is used there is still only one third of the total number of pixels available for each colour, this is the major drawback of single chip colour cameras.
3-CCD Cameras
For the best colour and image reproduction one CCD is dedicated to each colour. This is done by splitting the image into three and using three CCDs with primary colour filters to produce the respective primary colour outputs. This principle is illustrated below and is the method used by most 3-CCD cameras.
Dichroic filters are made from crystals which reflect light above or below a certain wavelength and allow the transmission of the remainder. The main difference to normal filters is that no absorption occurs.
The first prism is coated with a dichroic filter on the back that reflects blue light and allows red and green to pass through. The second prism reflects red light and allows the green to pass.
Using this method, each colour channel has a dedicated CCD and processing channel, so producing the maximum colour definition and resolution.
| RED + BLUE + GREEN | = | WHITE |
| RED + BLUE | = | MAGENTA |
| RED + GREEN | = | YELLOW |
| BLUE + GREEN | = | CYAN |
It follows then, that cyan is the complement of red, yellow of blue and magenta of green.
| CYAN + RED | = | WHITE |
| MAGENTA + GREEN | = | WHITE |
| YELLOW + BLUE | = | WHITE |
Optical effects, particularly unwanted ones, often refer to these complementary colours so it is as well to know their composition.
CCD Sizes
CCD arrays are available in a range of sizes. These are nominally, 1”, 2/3”, 1/2” and 1/3”. The term nominally is of significance as the sensor size is not physically accurate. The reason why the sensor size names are larger than the physical size of the array again goes back to the electro-mechanical limitations of early tube cameras. Tube cameras worked in a very similar way to a CRT in reverse. In simple terms it consists of a glass tube, with a rectangular plate at one end. This plate is coated with a photo-resistive material (e.g. lead oxide) and modulates the current of the scanning electron beam emitted from the cathode. Although it is the plate which equates to the image sensor size, the tube diameter is the figure used for specification.
Since this diagram was originally drawn, CCD sensors have increasingly become more miniature. It is now common to have 1/4” and 1/6” CCDs. But care should be taken. Lens production cannot keep pace with circuit miniaturization and so the more the sensor reduces in size, the more demands are placed on the lens to achieve the same light input and resolution.
Progressive scan, HD and widescreen
This process describe above of interweaving lines from alternate fields is called interlacing.
Some digital camera systems, especially HD, refer to a progressive scan process.
This means that the lines are sequential output from one frame to another. This offers better images since each line (and therefore pixel) is capturing the same moment in time.
However, some HD cameras offer more lines than standard definition, but still employ an interlacing technique.
Furthermore, HD images are usually not in the 4:3 ratio, as described above, but 16:9 or popularly known as widescreen. Therefore, video formats have grown from essentially a single standard definition image of 625 interlaced lines of video to quite a range, summarized below.
The table below covers UK and European television formats not those in the USA.
| Format | Ratio | Horizontal resolution | Vertical Resolution | Scan mode |
| SD 576i or 576p | 4:3 | 400-900TVL | 625 lines | Interlaced or progressive |
720i | 16:9 | 1280 | 720 | Interlaced |
720p | 16:9 | 1280 | 720 | Progressive |
1080i | 16:9 | 1920 | 1080 | Interlaced |
1080p | 16:9 | 1920 | 1080 | Progressive |
UHD | Under R&D |
HD is normally encoded as a digital MPEG2 signal, although some encode with MP4.
Displaying HD (HD Ready)
Screens that can accept HD signals must bear the “HD Ready” symbol or logo.
The EICTA introduced the logo to indicate whether the display can genuinely receive and display HD images, however, they relate only to DVI, HDMI and Component analogue inputs.
The label is used as a quality sign for the differentiation of display equipment that is capable of processing and displaying high-definition signals. It is awarded on the basis of minimum functionality requirements that are detailed in the "EICTA conditions for HD Labelling of Display Devices".
Warning: a display can bear the logo and not actually display the full picture resolution of the HD source. Some screens, CRT and plasma in particular, may only have a horizontal resolution of 1024 and so they cannot display even low HD 720p (1280 x 720). But they are HD ready since they can accept the input and provide a widescreen format.
This logo is more relevant to the consumer market and does not address the HD- format.
Many professional and scientific cameras will output on HD-SDI. This has two main modes HD-SDI and HD-SDI dual-link. HD-SDI will transport HD images at 720p and 1080i, but not, 1080p. For 1080p you require two HD-SDI lines, called dual-link.
Video Signals
Video output has four analogue forms, and an ever-increasing range of digital; the main types and uses are tabled below:
| Type | Signals | Uses |
| RGB | 3 | Used for high-end imaging. RGB is the best form of a live picture for imaging purposes. It maintains the maximum colour reproduction and is preferred by all good capture cards. |
| Component Y, R-Y, B-Y | 3 | A complex colour output format used for electronic film production (EFP) and electronic news gathering (ENG). It is the preferred format for broadcast tape recording. |
| YC | 2 | A semi-professional format that converts a colour image into a good quality monochrome signal (luminance) and outputs the colour information separately as a modulated signal (chrominance). This format is the preferred format for the S-VHS and low end DV recording formats. |
| Composite | 1 | A simple one-wire format preferred by the CCTV industries for easier wiring. The signal combines the luminance and chrominance signals together with synchronizing pulses (hence composite). The result is a moderate signal quality not recommended for imaging. |
| Digital | ||
| IEEE 1394 aka Fire Wire | 2 pairs optional 2 power | A serial digital video signal interface for interconnecting the DV (25 Mbits/sec) range of products and editing computers. It combines video, audio, MIDI and control signals and “sometimes” power (8-40V). However, it is not useful for long distances. |
| SDI | 1 | A broadcast quality digital serial interface, capable of transmitting the highest quality from the new professional digital formats over longer distances. |
| DVI-D | Multi | The digital form of VGA (data graphics)
Not useful for long distances. |
| DV | 2+2 | The recording format for IEEE 1394 transfer. 25 Mbits/sec 5:1 compression. Used in most domestic and more and more professional recorders. |
| HDMI | Multi | The successor to the DVI-I format, most commonly used for HD displays. Primarily designed for computer generated video and graphics but commonly used for home video. |
| HD-SDI | 1 | The HD version of SDI up to 1080i |
| HD-SDI dual link | 2 | The 1080p version of HD (also available as a single 3 gigahertz signal known as HDSDI3G |
Analogue Encoding | RGB
RGB, as mentioned, is the ideal format for imaging. It is the purest form as the image starts in RGB and is the form required for most output media. Capture cards ultimately require RGB separately as do display monitors (those without RGB terminals, decode to RGB internally). To synchronize the monitor or capture card to the video, sync’ pulses must be available, these are either on a separate signal (as per all of cameras with RGB output) or combined with the green channel (sync on green). |
It is important not to lose sight of the fact that all video cameras, evolved from broadcast television cameras ― and monochrome ones at that!
When broadcast cameras were first developed, they were monochrome, and the camera’s reaction to different colours was lost in the overall reaction to light. In other words, cameras could only produce an approximation to the lightness and darkness of a given scene ― this is called luminance.
As a result a solitary signal could transmit, by modulation at HF radio waves, all the necessary information to the home, where the TV would decode the incoming radio wave back into luminance.
When the idea of using three single tube cameras to produce RGB signals was developed, the resultant signal had to be compatible with existing, monochrome, TV sets. This meant that the modulated and decoded signal had to look like the original luminance one.
A typical luminance signal is shown below:
luminance of one TV line
Because of the impracticality of sending three or four signals to each home and the need to maintain compatibility, the RGB signals are encoded, this conveniently takes us through the other output formats mentioned earlier.
Component
The first task is to produce a luminance signal from the RGB signals. As the human eye is more sensitive to green and yellow than red and blue, different proportions of each are added together (59% green, 30% red and 11% blue). This combined signal from RGB is called luminance, is given the symbol Y and is identical and compatible with the original monochrome signal.
Adding Colour
To obtain colour information, all that is required is the difference between Y and each of the R, G and B signals. As G is the dominant component in Y, the signal G minus Y (G-Y) would be quite small and prone to noise. Therefore, only R-Y and B-Y are used, because if you have Y, R-Y and B-Y you can derive G:
Proof:
R+G+B = Y, therefore G = Y - (R+B)
If B = (B-Y) + Y and R = (R-Y)+Y
then G = Y - (((R-Y)+Y)+((B-Y)+Y))
= Y - (R-Y) - Y - (B-Y) - Y
= -Y-(R-Y)- (B-Y)
(the three signals)
The signals R-Y and B-Y are known as colour difference signals and together with the luminance signal (Y) they constitute the output format known as Component.
This is the first stage of the single wire encoding process but is often used for professional recording as it maintains good picture quality.
YC
The next stage of the encoding process is to reduce the two colour difference signals to one signal, this is called chrominance and is given the symbol C. This is encoded by altering (modulating) the amplitude of a high frequency clock with the two colour difference signals.
This is a very simplistic view of the colour modulation process. The reality involves signal weighting, quadrature modulation, phase relationships and sub-carrier suppression. Detailed descriptions are available from many good books on TV technology.
The encoding and subsequent decoding of this colour information degrades the colour content of the picture quite considerably. It is worth noting at this point that the human eye is relatively insensitive to colour compared with contrast. This is because only the cones are sensitive to colour and both rods and cones contribute to lightness (cones for bright light and rods for dim).
The encoding process uses this fact and so attempts to maintain the highest possible luminance resolution whilst accepting a rather poor colour reproduction. This may be acceptable for broadcast television but may not be for imaging ― for example, colour print registration, colour analysis, image analysis and output to pseudo-photographic print media, or just simply high quality images.
The result is a two-wire output format used by semi-professional VCRs, some monitors and a few image frame grabbers.
Composite
To complete the process of reducing the colour signal to be compatible with the original VBS monochrome composite signal, the chrominance pulses are added to the VBS signal in a redundant area called the back porch.
The colour information is contained in a short burst of pulses at a frequency of 4.43MHz. If you remember from the explanation on resolution in the section Units and Specification, an average camera would have 400 TVL of resolution in 3/4 of the width, this equates to 5 MHz. If you imagine a pattern that equated to 350 TVL, this equates to 4.4MHz, the same frequency as the colour burst. As a result such patterns create havoc with the colour reproduction. This factor, coupled with poor colour resolution and a host of other interference effects, renders composite video wholly unsuitable for imaging or quality video purposes.
Most cameras, however, provide an additional permanent composite video output for the purposes of monitoring and viewfinding.
Digital Encoding
All digital encoded formats have one thing in common. They are a representation of an analogue signal into a digital stream or data block.
From the perspective of the typical readers of this booklet, the importance is in the resolutions and quality of the signals they can transmit, the distances they can transmit over (unboosted) and the connector types employed. The best way to illustrate this is with a table:
| IEEE 1394 Fire Wire | 2 pairs optional 2 power | A serial digital video signal interface for interconnecting the DV (25 Mbits/sec) range of productsHowever, it is not useful for long distances. |
| SDI | 1 BNC | Am SD broadcast quality digital serial interface, capable of transmitting the highest quality over long distances. |
| DVI-D | Multipin | The digital form of VGA (data graphics)
Not useful for long distances. All HD formats. |
| DV | 2+2 | The SD recording format for IEEE 1394 transfer. 25 Mbits/sec 5:1 compression. Used in most domestic and increasingly more professional recorders. |
| HDMI | Multi | The successor to the DVI-I format, most commonly used for HD displays. All HD formats over short distances. |
| HD-SDI | 1 | The HD version of SDI up to 1080i (1.5Gbits/sec) |
| HD-SDI dual link | 2 BNC | The 1080p version of HD-SDI over two co-axial cables. (to 3gbits/sec) |
Light
Any system is like a chain and, thus, is only as strong as its weakest link. To get the most out of any video imaging system, a sound knowledge of optics is essential. This should range from lens theory to light qualities and properties. This section deals only with the basic subject matter.
The first link in the imaging chain is the lighting conditions. It is essential that the subject is illuminated to the best effect. This can involve simple tasks, such as shadow control or more complex solutions, such as the use of polarized light, light of a controlled bandwidth and wavelength in conditions where good lighting is impossible or difficult.
White Light
White light is the combination, in equal quantities, of all the visible wavelengths of electro-magnetic radiation (between 400 and 780 nm). When analyzed in video terms, it is the combination, in equal proportions of red, green and blue signals. However, when the illumination is with light other than white light (practically always) some adjustment to the “recipe” is necessary.
White light, defined for video as being “standard daylight”, is a mixture of direct sunlight and north sky light. This is sometimes referred to as illuminant D.
Colour Temperature
The need to measure “whiteness” led to the adoption of a colour temperature scale. This equates a certain colour of light to the temperature that a “black body” must be raised to produce light of the same bandwidth: “Standard daylight”, for example, has a colour temperature of 6500°K. Consider the case where a sheet of paper is illuminated with a tungsten lamp. The camera would output considerably more red energy, but the observer still wishes to see white. To facilitate this, the camera will increase the temperature of the scene by reducing the red and increasing the blue levels. This adjustment to the RGB channels is called white balancing.
Below is a graph illustrating sunlight against a tungsten lamp and some common light sources with their respective colour temperatures, measured in Degrees Kelvin (°K).
| Light Source | Effect | Colour Temp |
| Incandescent lamp | very red tint | 2850 |
| Halogen lamp | slight red tint | 3200 |
| Fluorescent lamp | mainly white | 4500 to 5500 |
| Sunlight | mainly white | 5500 to 6500 |
| Illuminant D | white | 6500 |
| Cloudy sky | blue tint | 7000 to 8000 |
| Xenon flash | blue | 9500 |
A word of caution, if the light used is at either extreme then the extra amplification required to the red or blue channel can introduce considerable noise (see section on signal to noise). In these cases it is better to colour correct using optical colour filters.
Coloured Light
The bandwidth of visible light is between 390 nm and 780 nm. Fortunately, if two or more rays of different wavelengths arrive in close proximity on the retina, the human brain interprets this as one ray of the average wavelength. Thus by varying the amplitudes of red, green and blue, most of the colours of the spectrum are reproducible.
Colour | Wavelength | Frequency | RGB Mix | ||
| nm | 10 | 12 HzRed | Green | Blue | |
| Red | 780-622 | 384-482 | 100% | - | - |
| Orange | 622-597 | 482-503 | 80% | 20% | - |
| Yellow | 597-577 | 503-520 | 50% | 50% | - |
| Green | 577-510 | 520-588 | - | 100% | - |
| Cyan | 510-490 | 588-611 | - | 40% | 60% |
| Blue | 492-455 | 610-659 | - | - | 100% |
| Violet | 455-390 | 659-769 | 40% | - | 60% |
Violet is approximated by magenta, which is the combination of red and blue (the complement of green). It is a non-spectral colour and magenta strictly has no wavelength associated with it.
Hue, Saturation and Brightness
Often, image analysis and image capture software refer to colours as hue, saturation and brightness, (HSB), this is sometimes known as HSI, with I being intensity.
Hue
Different colours are made by different ratios of R,G and B, this is known as hue. It is sometimes described by a colour triangle or wheel. The wheel can be thought of as having angular properties and hues are often given degrees of rotation.
Saturation
Paler shades of a given hue are achieved by adding the complement in varying degrees. For example, if you take red and add a little cyan (blue and green), the red will appear thinner or paler, tending to white. This is altering the saturation level, i.e. pink is less saturated than scarlet, but pink and scarlet have the same hue.
Brightness (Intensity)
The overall brightness is controlled by altering, proportionally, all RGB levels together. RGB all at 100% output gives white, anything less is grey increasing to black. The same is true for any hue and saturation.
Lenses
focal length, f
The second link in the chain is the glass used to focus the image onto the sensor. A simple lens uses the fact that glass bends light depending upon the angle of incidence. In a convex lens, this has the effect of converging light rays behind the lens and so focuses images onto an image receptor, e.g. CCD.
It is fairly obvious that the lens must be clear and smooth. The smoothness and accuracy of curvature, together with the optical quality of the glass are the factors that determine the resolution of the lens. This is why a 50mm lens can range from a few pounds to a few thousand pounds. It is a common mistake to assume a lens is a lens and this assumption all too often leads to a poor imaging system. It is not the intention of this booklet to detail comprehensive lens theory but to illustrate the major problems and limitations.
Resolution
The resolution of a lens is not often quoted. However, if provided it will be given by its Modulation Transfer Function (MTF). This is not too dissimilar to a camera’s horizontal resolution specification given in TV lines. As mentioned in the camera section, one cycle of a given frequency represents one black and white pair of lines. Lens specifications refer to these as line pairs per millimetre. Unlike a camera, this has no time factor and is referred to as spatial frequency.
As the spatial frequency increases, the more difficult it is for the lens to resolve them. What happens is exactly the same with cameras. The peaks of the white lines merge with the trough of the black lines until they become extinct.
The MTF is usually shown as a curve for a range of frequencies.
100%
The MTF actually varies for different zones , the central part being generally the best. This gives rise to the mistaken notion that a 2/3” lens is better on a 1/3” camera than a 1/3” lens. The reality is that the 2/3” lens must have twice the maximum spatial frequency (lp/mm) if the image width is halved.
Note: The MTF of a given lens is affected by aperture, change in focal length (for zoom lenses), the distance from the object and the lighting conditions.
Focal Length - f
This is the figure that determines the angle of view of a lens. It can be shown that the ratio of image width to object width is the same as focal length to object length.
From a CCD point of view, the maximum image width is known (size of sensor), the focal length of the lens is known. Therefore, using the formula:
| Maximum image width | image length | |
| ___________________ | = | _____________ |
| Maximum object width | Object length |
Basic trigonometry gives the angle of view as :
Angle of View = 2 x tan-1 (maximum object width/object length)
Useful Tip: The formula can show that an object at infinity will be focused at a distance equal to the focal length. Therefore, to get an estimate of the focal length, focus the image of a distant window onto a wall.
FOV = 2 x tan-1 (sensor width/2 x focal length)
The table below shows the angles of view for a range of common lenses.
Focal Length, f | Angle | of View | ||
| mm | 1/3” CCD | 1/2” CCD | 2/3” CCD | 35 mm |
| 5 | 51º | 65º | 82º | 148º |
| 15 | 18º | 24º | 33º | 100º |
| 25 | 11º | 15º | 20º | 71º |
| 50 | 5.5º | 7.3º | 10º | 40º |
| 100 | 2.7º | 3.7º | 5º | 20º |
Aperture
This is the diameter of a lens. It is often controlled by a diaphragm to restrict light and to alter the depth of field. Large apertures cause blurred images due to aberrations and very small apertures cause blurred images due to diffraction (see explanation at the end of this section). It is important, therefore, to maintain a lens system in the mid-range of its aperture.
F-Number
This is a measure of the aperture setting. The F-number is simply the ratio of focal length/aperture:
F-number = f/A
This is sometimes referred to as the speed of an optical system including cameras. Most camera systems are limited only by the lens, which is why cameras are tested for sensitivity at the widest aperture F1 or F1.4. What you should be aware of for good imaging is to maintain the aperture setting between F4 and F11. This is where the sharpest images will be obtained provided there is sufficient light.
Depth of Field
This is a function of the aperture. At large apertures the depth of field is very shallow and at small apertures the depth of field is deep. Depth of field is a function of the hyper-focal distance. This is a point from the lens where everything beyond is in focus and is given by the formula:
h = f2/c x F-number
where, h = hyper focal distance
c = circle of least confusion
and f = focal length
As can be seen from the formula, as the F-number is increased the hyper-focal length draws nearer, hence the depth of field increases. However, as already mentioned in the paragraph on aperture, if the aperture becomes quite small (F16 or F22) diffraction will cause blurring of the image.
The depth of field for a given object distance can be calculated from:
depth of field = T2-T1
where T1 = B(h+f)/h+B for near limit
T2 = B(h-f)/h-B for far limit
and B = object distance
Explanation of Diffraction
Diffraction is the special name given to a type of interference. It is caused by waves (light) changing direction after striking the boundary of an opaque object (this accurately describes the rôle of an aperture). Light which passes close to the boundary is caused to change direction, the amount deviated depends on the wavelength of light. It follows then, that the smaller the aperture, the more of the image is formed by such rays, giving rise to a blurred image with coloured fringes.
Optical Problems
Lenses are far from perfect. The property of transparent material that allows lenses to work is, ironically, the reason why they do not work well!
Lenses work because they change the direction of light a varying amount, depending on the angle of incidence.
Chromatic Aberrations
The problem arises with light of different wavelengths. The angle of deviation not only depends on the angle of incidence but also on the wavelength, with blue being bent the most and red the least.
Snell’s Law is thus modified for the refractive index at a given wavelength. There is no easy formula as different materials behave very differently from each other. The absolute refractive index quoted above is for yellow light unless stated and assumes air to be the same as a vacuum, i.e. n=1.
This gives rise to two errors, axial and lateral chromatic aberrations. Axial being the defocused image of red relative to blue (or vice versa) and lateral being the effective larger magnification of the red light (longer focal length) on the image surface. To minimize these effects, glass of complementary refractive indexes are bonded to correct for these errors. These are known as achromatic lenses.
Note: The quality of this chromatic correction is reflected in the price of a lens.
Spherical Aberration
As rays strike the lens further away from the centre (optical axis) the focal length moves towards the front face of the lens (vertex). This effect causes soft or defocused images at wide apertures. Stopping down the lens greatly reduces this effect but has the disadvantage of reducing the light transmission.
Spherical aberrations are always present in lenses but are minimized in well-designed multi-element lenses. Needless to say, this is also reflected in the price.
Comatic Aberration
The ideal lens is wafer thin. This is not practical in reality and they are always made from relatively thick glass, mainly to accommodate the radius of curvatures required and for strength. The result is a curved effective refractive plane causing a point of an object to be focused at different places on the image, depending on the path taken by the rays.
Astigmatism
In the ray diagrams used, all rays are shown in 2-D on a plane containing the optical axis. These are called meridonial planes. Rays also travel on an infinite number of planes that do not contain the optical axis. These are called sagittal planes. A ray from a point off-axis is more angled on a meridonial plane than the corresponding sagittal plane. Thus the two rays have different focal lengths, the meridonial plane being shorter.
This error is entirely relative to the distance of an object from the optical axis. It is not improved with aperture reduction.
Field Curvature
Any lens only perfectly focuses concentric object planes onto concentric image planes. Unfortunately, when the object is flat (giving variable distances from the lens face), the best image plane would be even more curved. On a flat image plane, e.g. CCD, this causes a lack of focus in the corners
Distortion
Different areas of the lens have different focal lengths and therefore, different magnifications. This gives rise to the last type of aberration, distortion. There are two types.
For imaging purposes most lenses are convex and cause barrel distortion. Severe barrel distortion is typical in wide angle and fish-eye lenses, as the distortion is somewhat proportional to the distance from the optical axis. The position of the aperture can also affect the nature and magnitude of the distortions and is often used to correct distortion from lenses.
Lens Types
Fixed Focal Length
These lenses are of a finite focal length and the elements, glass and aperture positions are designed to optimize the performance for that focal length. Fixed focal lenses offer the best performance for a given focal length.
The minimum working distance can be reduced by the use of spacers (extension tubes) to optimize a system for a given application without any degradation to the lens performance. This is very useful in fixed format imaging.
Variable Focal Length
This type of lens is one that has a range of focal lengths. Different focal lengths will alter the different focal plane for a given image. In other words a change in focal length will require the image to be refocused.
Generally, the user selects the focal length required and secures this by the use of a thumbscrew. The lens is then treated as a fixed focal lens. Because of the moving parts, the performance at a given focal length will probably not be optimized as with a fixed lens.
Zoom
A special kind of variable focal lens. At a certain distance behind the lens, the image is in focus irrespective of the zoom position (focal length), this is called tracking. On any other plane, the image would need refocusing. It is important, therefore, that the CCD is placed at this location. This is done by altering the back-focus, i.e. moving the CCD back and forth, to focus the CCD to the lens.
Close-up Dioptre
This adaptor makes infinity appear nearer the lens. The distance away depends on the dioptic power, and is given by: 1000mm/Dioptic Power.
Therefore. +1 = 1000mm, +2 = 500mm +3 = 333mm. Because this is a lens like any other, it will suffer from all the errors mentioned earlier and the better types are achromatic. Beware of single element dioptres.
Lens Summary
All of the aberrations and distortions mentioned earlier can be compensated for by careful positioning of the aperture and by the introduction of elements having the inverse property to the error to be corrected. This is a complex matter involving lengthy calculations and the use of many elements often of different optical glass. For imaging it is essential to use good lenses that have the following properties:
a. the right image format e.g. 1/2” lens for 1/2” camera.
b. comparable resolution to the camera used and to the image output requirement.
c. low distortion.
d. the least amount of monochromatic and chromatic aberrations.
Hopefully you begin to appreciate the importance of a lens in the system especially for scientific or professional applications.
It is too easy to buy into the “one lens fits all” philosophy of marketing.
“10x optical zoom, 6 x digital zoom, macro setting for close up, etc.”
Yes, of course, it will produce an image and for holiday snaps probably acceptable. But ask yourself why professional photographers still carry large body cameras and have a range of lenses with them that cost thousands of pounds.
Well part of the answer is described in the complex mathematics of trying to control light.
A thought:
Try not to imagine light as being one colour, i.e. “white”. Imagine it as 16 million colours interacting. Then imagine working out a lens system to keep 16 million variables under control.
You begin to understand the problem?
Resolution
A whole section has been dedicated to this subject because, as was already mentioned, it is an area of imaging that is often confusing and misunderstood. It is also affected by almost every aspect and component in a video system. Each aspect is dealt with separately here.
Movement
It stands to reason that if an object is moving it cannot be resolved as well as a stationary object. This situation is fairly obvious with respect to people and cars. However, vibration, even if humanly imperceptible, can have a devastating effect on resolution. Slow scan cameras are particularly prone to this cause of loss of resolution.
Light
Because of the chromatic aberrations mentioned in the Lenses section, it follows that if red light is focused at different places to blue light, a scene which is illuminated only with red or blue light will render better resolution. In many instances this is not practical but it is worth bearing in mind that if monochrome images are what is required, then the resolution will be improved by using only a narrow spectrum of colour, say green.
Lens Resolution
As mentioned in the Lenses section, the finite resolution of a lens is a complex function and dependent on many other factors, such as, aperture, distance, angle of view and focal length. It is also the case that a lens which can resolve high spatial frequencies, necessary for ENG and EFP use, often perform less well at the lower frequencies.
See the paragraph on Choosing a lens in the Systems section.
Camera Resolution
Camera resolution is dealt with under specifications and units. However, much like the imaging system being analysed in this booklet, the camera is a also a system and resolution is affected by the component parts. These are fundamentally the horizontal pixel count, any filters in front of the CCD, the internal optics (e.g. prism block for 3-CCD), any internal lens (e.g. sensor reformatting lens), the quality of electronics and amplifiers (bandwidth of components), signal processing, digital signal processing (number of bits) and the quality of encoding circuits.
The number of pixels in the horizontal only relates to a theoretical resolution. Techniques to recover lost resolution are covered in Picture Recovery.
The vertical resolution of cameras is normally lower than the horizontal. This is due to camera manufacturers’ striving for sensitivity. Particularly in still imaging, it is common practice to add two lines of the CCD together and output them as one, be it a monitor, recorder or a capture card. Many cameras can operate in single line mode, improving the vertical resolution by 40%.
Capture Card
If a frame grabber or video input card is in the system, then its resolution will naturally affect the overall. Unlike a camera, the resolution is quoted in pixel counts as it more accurately reflects the resolution, but any decoding at the input stage (from composite to RGB) will have an adverse effect.
A common pixel count is 768 X 576 for good quality capture cards. This equates to square pixels for 50Hz based video (PAL) and is considered a standard digital PAL frame. The number of pixels wide is controlled by a pixel clock and available memory.
A good video card or frame-store will grab images at the quoted resolution without noticeable degradation. It does not suffer from the blurring problems of optics or analogue signals, but does suffer from the quantization errors of DSP. Harmonics and moiré patterns are often more problematic as the capture card will tend to see “clearly” only some of the structure observed and freeze the unwanted pattern.
Monitor
CRT
If the image is to be viewed on a monitor, then the monitor has a limiting resolution to contribute to the system. A quality 3-CCD camera can only really be appreciated to the full on a monitor of Grade II quality. A monitor’s resolution is quoted in a similar way to a camera’s (TVL). In a colour monitor, the resolution is fundamentally dictated by the effective trio-dot pitch, this is 0.28mm on good quality grade II monitor. It should also be noted that the quoted resolution of a monitor is at extinction.
By rough calculation:
a 14” CRT is 11.2” wide (8.4” high)
this width is 285 mm divided by 0.28 dot pitch = 1016, say 1000
If you assume one cluster (trio) of dots for one line, then this equates to 1000 lines or 750 TVL, measured in the central 3/4 region.
LCD
More and more likely, the image will be viewed on an LCD screen. These screens quote native resolution and that is the important part. These screens are not raster screens as in CRT and basically attempt to map each output pixel from your image device, to each display pixel.
So it follows that you should pick a native resolution that is equal or greater to your minimum resolution in your system. For example, if you have a 1080p camera (and we are assuming that by now you have invested in a decent lens), then make sure you have a true HD screen with a native resolution of at least 1920 x 1080 pixels.
Interestingly though, a screen with a higher resolution may not produce sharper images, why? Well imagine you have a 720p camera giving you 1280 pixels. To show this on a true HD screen with 1920 pixels, the screen has to interpolate the missing pixels. So for every two pixels output, three are used to display.
From camera χχχχχχ
To screen χχχχχχχχχ
This means the first pixel is spread across pixels 1 and 2, the second pixel spread across 2 and 3, the third pixel is across 4 and 5 and so on. This means that every third pixel is effectively “made-up” or created by averaging the respective input pixel values. Under certain conditions, this can create erroneous displays, particularly for still images.
You may find a sharper image on a native 1280 screen.
Analogue VCRs
Apart from archived material, this format should be obsolete. Tape recording normally inflicts the most damage to a video image. This is because of the magnetic nature of recording and the use of encoded signals. Generally VHS recordings reduce the resolution to 240 TVL and S-VHS can only maintain 400 TVL.
Digital VCRs and DVD recorders
Digital VCRs offer higher resolution recording but the compression of the various formats is the major contributor to loss of resolution.
Most low-cost commercial tape recorders are DV and you should refer to the earlier section on DV to understand the resolution. However, these do represent fairly lossless recordings and are a much better solution to video recording than the above, which you should avoid at all costs. For standard definition, 768 x 576 individual frames can be stopped or paused and accessed as bitmaps maintaining good resolution.
DVD recorders normally employ MPEG2 encoding at a fairly decent compression rate.
Most offer long play or extended play due the shortage of memory on a disc compared with a tape. Care should be exercised here, since the long play recordings effectively record, well less! Less frames, less pixels and more compression equals poorer images.
Printers
The resolution of a printer will depend mainly on the type of process used and the quality of paper printed on. For imaging purposes the best resolution is with a dye-sublimation printer. However, other types exist such as, thermal wax, laser jet, ink jet, bubble jet, dot matrix and thermal paper. All printers will quote resolution as Dots per Inch (DPI). What is often confusing is how do DPI equate to TVL or pixels.
Below is an illustration of how they approximate to standard paper sizes.
| 75 DPI | A4 (11.7 in max.) | 877 pixels |
| 150 DPI | A5 (8.3 in max.) | 1245 pixels |
| 150 DPI | A6 (5.6 in max.) | 840 pixels |
| 300 DPI | A4 | 3510 pixels |
| 300 DPI | A5 | 2490 pixels |
| 300 DPI | A6 | 1740 pixels |
Dye-sublimation versus inkjet?
A common mistake is to compare the DPI of an inkjet or laser jet to a dye-sublimation.
This is completely erroneous and would lead an unaware buyer, incorrectly comparing resolutions and assuming that a 1400 dpi inkjet is better than a 300 dpi dye-sublimation.
A dye-sublimation has the unique property of being able to place any colour onto a single dot (the colours melt into each other in the substrate of the special paper). Inkjets and deskjets make up colour by arranging a pattern of dots with colours from the wells that they contain, usually from red, green, blue, black, magenta, cyan and yellow.
For example green could be from an array of dots of yellow and blue, immediately, halving the resolution. A very light shade of green would incorporate more white space, perhaps dividing the resolution by 10, for say 10% saturation. This would have the effect of making that part of an image printed on a 1400 dpi inkjet approximately 140 dpi, less that half that of a 300 dpi dye-sublimation.
The unfortunate thing is that flesh tones are a complex colour for inkjets and deskjets and all too often are the areas that demonstrate the limitations of this process. Dye-sublimation on the other hand is always at the same resolution, whatever the colour.
Human Eye
A factor in any overall imaging system, often not considered, is the resolution of the human eye. The average human eye is capable of resolving one minute of arc (1/60th of a degree).
To give an indication of how this relates to typical image output media, refer to the table below:
| Output | Distance |
Standard 6x4 photograph.......................... | 15 cm |
300 DPI print............................................. | 30 cm |
150 DPI print............................................. | 60 cm |
75 DPI print............................................... | 1.2 metres |
800x600 14” SVGA Computer Screen...... | 1.2 metres |
1024x768 17” XGA Computer Screen…... | 1.2 metres |
1024x768 14” XGA Computer Screen.... | 1.0 metres |
750 TVL 14” monitor.................................. | 1.0 metres |
400 TVL 28” television............................... | 2.6 metres |
Systems
The first task in building a system is to determine what is the object of the imaging system. This should give an idea of required resolution and colour reproduction. For general viewing, surveillance, CCTV installations etc, single chip cameras of moderate resolution are acceptable. This section will deal with more demanding requirements and lead through each aspect.
Motion Control
To handle moving subjects, there are a variety of different imaging techniques. The most common is to control the exposure time for each frame of video. This is known as electronic shutter and is analogous to the mechanical shutter of a conventional film camera, i.e. the faster the shutter the crisper the moving subject. Most imaging cameras can shutter in fixed steps to 1/10000 second (100 microseconds). Care must be taken, however, as this has the effect of dramatically reducing the amount of light reaching the sensor.
Some high quality imaging cameras can shutter up to 1/10000 in small increments to match repeating images (e.g. computer screens). They can also be synchronized with external signals (random trigger) to capture a moving image, or a rare event, at the right time.
Another method is to use a stroboscopic light, this has the same effect as shutter but is capable of operating at much shorter exposures, typically a few microseconds.
Finally, a new technology of using LCD shutters is available. These can operate at much faster shutter speeds than the camera but without the need to control the ambient lighting.
Lighting
Careful consideration should be given to the lighting at all times. Colour temperature and white balancing has already been mentioned, but it is good practice to try and use filters in extreme cases. For example, if your light source is very blue e.g. xenon, then a red correction filter will restore the balance optically without the need to boost artificially the red content, risking the introduction of red noise. Similarly, a red light source, e.g. an incandescent lamp on low power, would be best corrected with a blue filter.
These are extreme cases and only the better imaging cameras have the range to cope with normal light selection (see Colour Temperature).
Another lighting problem can be the frequency with which certain lights work, e.g. fluorescent lamps. Because the frequency of the lamp is close to the scanning frequency of the camera, a beat frequency (harmonic) can sometimes be observed.
Some cameras have a 1/120 second normal shutter option to remove this effect. Other cameras have a sensor, which detects fluorescent light and compensates automatically with a shutter speed change.
.
Choosing a Lens
The focal length and type of lens is determined by first calculating the field of view (or range) required as shown in the section Lenses. The next considerations are the resolution required, the minimum object distance or working distances and the quality of lens.
If the application requires good performance in all zones of the resultant image, then a quality lens with good chromatic correction and the required MTF must be selected.
An example:
A rough calculation to select a lens suitable for a 1/3” 3-CCD camera.
Camera Specification:
sensor width is 5mm (1/3” CCD)
ideal resolution 750 TVL
750 X 4/3 = 1000 full screen width (therefore, image width)
1000 ÷ 2 = 500 convert to cycles or line pairs
500 ÷ 5 = 100 divide by sensor width
lens resolution = 100 line pairs per mm.
The true MTF of a lens is best demonstrated by graphs. Below are typical curves for different lenses.
Initially, lens 2 may seem a better choice since it performs into the 150 lp/mm area. However, converting the spatial frequency to TVL and superimposing the cut-off resolution of the other system components can be more informative.
Choosing a Microscope Adaptor
Most modern compound and stereo microscopes are available with a trinocular head where a video output is required as well as the basic binocular eyepieces. Originally the optics involved were designed to project the image onto a 1” tube camera allowing simple mechanical adaptors to interface the camera with the microscope. Unfortunately, the camera sensor size reduced but the port for the cameras stayed with the old format 1”. Hardly any 1” cameras are used now and optical adaptors are usually required, particularly for 1/2” and 1/3” cameras.
The image circle is 16 mm in diameter and in the case of the 1/3” camera; there is a 3X unwanted magnification. This is rectified by optical adaptors, which resize the image from 16mm to the appropriate sensor format.
Adaptor | Magnification |
| 1/3” C-Mount | 0.37X |
| 1/2” C-Mount | 0.5X |
| 1/2” Bayonet | 0.5X |
| 2/3” C-Mount | 0.69X |
Shading
This is an appropriate point to discuss the phenomenon of vertical shading. This can present itself in any optical system and it is an inherent property of any prism camera. As such, this point is highly relevant to 3-CCD cameras. The cause of vertical shading is the differing wavelengths of different coloured light.
As was mentioned earlier in this booklet (see Optical Problems), a ray of light is refracted an amount depending on the angle of incidence and the wavelength. In the same way that lenses are not perfect, dichroic filters reflect light of various wavelengths, depending upon the angle of incidence, see diagram.
lens
From the diagram it can be seen that the angle of incidence is more acute at the bottom of the filter than at the top. This changes the wavelength that is reflected. In the case of the blue separation filter, the more acute the angle of incidence, the more green is reflected with the blue. In extreme cases this manifests itself as a cyan to magenta shift vertically down the monitor screen.
Care must be taken to choose a camera that allows for adjustment to trim out this unwanted shading. Correction is by saw-tooth adjustment in the vertical.
Capture Cards
Image capture cards are also known as frame stores, frame grabbers, digitizers, image grabbers or basically any two words that mean video picture and capture. They are fundamentally different from image processors and from image analysis packages that thankfully do not have so many pseudonyms and which only deal with images in their post-capture state.
Essentially, a capture card is an electronic circuit board that slots into a compatible computer and has a video signal input and on-board processing to enable image acquisition.
Inputs
For imaging it is important to ensure the capture card has a multiple pin connector to enable RGB, YUV or digital video inputs. The resolution implications are explained in Video Signals.
Many good capture cards will have a trigger input to synchronize with external events, this leads to a very interesting application with slow-shutter cameras, see Picture Recovery.
Processing Power
Most frame grabbers are dedicated single board computers. As such most of the electronics are digital in nature and the performance measured by number of bits, data transfer rate etc.
The standard interface medium for computers is called a bus and the more recent PCI bus is common to both Apple and IBM and is the first cross-platform bus. Any new system being assembled may well consider this bus as it is extremely fast, not only allowing the use of the host memory and so keeping costs down but also providing real-time refresh rates (50Hz).
Resolution
The resolution has two aspects, pixel count and colour depth. The pixel count was covered in the section on Resolution. Colour depth is of equal significance. Each channel is converted to a digital form via an Analogue to Digital Converter (ADC). This must be very fast and have a sufficient number of bits to output in digital form. 8-bits, which is a typical figure, will provide 256 levels. Normally, a capture card will quote the colour depth as the total number of bits available to each colour, so as there are three colours, 24 bits would be typical. It is strongly recommended for professional colour imaging that a 24-bit capture card is used as a minimum standard.
24-bits (8 per colour) computes to 16 million colours. This is largely held to be the limit of the most discerning human eye. However, the eye may not be the overriding factor, for example, in a 10-bit per channel, colour analysis system, a 30-bit card with 1 billion colours would be necessary.
Output
The output from a capture card is normally viewed on a computer screen by using the computer graphics capability and the software picture in window modes. It is important to note that although most computer systems are supplied with capable monitors, they do not normally come supplied with an adequate graphics card to provide a display.
Below is a table outlining the different modes of display and the memory requirement.
Mode | Number of Colours | Memory for 768x576 pixels |
| 8-bit | 256 | 442 Kb |
| 16-bit | 65,536 | 884 Kb |
| 24-bit | 16,777,216 | 1.32 Mb |
Choosing a Display Screen
It is normally necessary to view the live image on some form of CRT or LCD display. In image capture applications this could mean a computer screen or video monitor. There are pros and cons for both.
Video monitors tend to give a higher quality of colour image, with good colour saturation and contrast, whereas, computer displays, although improving, are regarded as being “thin”. Certainly, for live imaging, particularly in respect of colour reproduction, a separate quality colour video monitor is preferable.
On the other hand, computer monitors are able to display higher resolutions by virtue of a closer dot pitch with less energy (ergo, smaller spot sizes) since they are intended for close viewing (less than 1 metre). This, together with their non-interlaced display modes, is a great benefit for still imaging at close range.
Computer Monitor versus Video Monitor
Video monitors have similar design parameters to TV or HDTV, that is, they are intended for viewing from a comfortable distance, say one or two metres, as true a reproduction of the scene as possible. To do this they must be bright, of high contrast and with strong colours. To accommodate these requirements, a lot more phosphor per dot is required to achieve a brighter output.
This restriction on the minimum dot pitch reduces the smallest spot size, exacerbated by the fact that more energy is beamed at the spot causing more flare.
LCD or CRT
There is little choice now since traditional CRT production has all but stopped. The important thing about choosing an LCD screen is to consider viewing distance and viewing angles, brightness and resolution.
Simple mathematics (illustrated above) will allow you to pick the screen you need. For video images, the optimum distance is 3-6 times the screen width, so:
Example 1
For a single viewer one metre away, a screen of about 15” is fine. Viewing angle of the screen is less important that maximum brightness (always from direct).
Example 2
For a room of a small group , say twelve people, where the average distance is 3 metres to the screen, then a 42” screen would work well. Viewing angle paramaters are vital, since those on the flanks may not see well if the screen has a narrow viewing angle.
The drop off of intensity per angle is specified on the product information.
Example 3
For a lecture theatre of 200, with a maximum distance of 15 metres, then a screen of 2 metres would be required with an appropriate resolution, high power projector.
But viewing angle is important here too, and so the screen surface must be carefully considered. High performance screens (designed to compensate for low power projectors) are very directional.
Picture Recovery
Due to the fact that a video signal goes in and out of so many circuits, some degradation to the signal is inevitable. Signal processing can be performed to recover this degradation on analogue signals, as in conventional cameras, or on digitally managed signals, as in DSP cameras.
The critical factor in respect of image quality is the sophistication of the signal processing and not whether it is DSP or analogue. The latest trend for DSP was brought about by designers utilizing the latent processing powers of digital circuits that already existed. It could be said that the greatest advantage of DSP is for the manufacturer.
Contour Enhancement
With the signal being input and output through various electronic stages, it is quite normal for some rounding of the corners to be experienced. This has the effect of being defocused, consider the square wave of the signal below before and after.
The amount of enhancement used is controlled by the contour function to be found in all high-end imaging cameras. There are 200 levels of enhancement and the amount required is dependent on the composition of the scene. For text it is important not to have too much enhancement as this causes artificial black or white shadows.
There are two types of contour enhance, 1H and 2H contour. 1H contour gets its name from the fact that it analyses each horizontal line as it is scanned, that is one horizontal line. The drawback is that under certain conditions, an edge is not detected until it has already passed by, this can give rise to non-symmetrical enhancement, particularly on rectilinear subjects, such as text. The better process is 2H contour, this requires the use of a delay line to analyze two horizontal lines, the current and the previous. This delay enables symmetrical enhancement. 2H contour is a prerequisite for the broadcast industry and is the preferred process in imaging.
Gamma
Unlike contour enhancement, gamma correction is needed to make up for the shortcomings of CRTs. The problem is that whilst CCDs are very linear (that is, the amount of charge stored is directly proportional to the amount of light received) CRTs are not. They are very inefficient at the low end resulting in a loss of detail in the dark areas of a picture.
The curve for the CRT, in mathematical terms, is close to a XY curve. The term Y is referred to as gamma and has the symbol γ (Greek letter for gamma).
The value of gamma is around 2.2. To correct for this, the camera signal is processed by an amplifier with the inverse curve, typically of 0.45 gamma value.
CCD sensor
The gradient of the curve shows compression at bright levels. This is good for highlight compression e.g. of car head lamps or of specular energy from metals or jewellery, but is not good for true image reproduction.
This has the adverse effect of compressing the information in the bright regions, losing dynamic range ( the gradient is shallower at the higher end of the light scale).
This normally manifests itself as a loss of detail in bright scenes. If you point the camera at clouds, with gamma on, there will be little detail in the clouds and often, they will be indistinguishable from blue sky.
At the outset, video dynamic range is desperately short compared with real life (when was the last time you had to squint at your TV set?). Gamma should therefore be selectable and used to get the most out of the image. When using a capture card, try to keep gamma correction off, as the capture card will then have linear signals to digitize, giving more accurate readings for colour saturation levels etc. Gamma correction can then be carried out post capture, to best suit your output format.
Master Black
Master Black is a useful technique of altering the black level of a video signal. To explain this, it is best to cite two examples:
If you want to grab text and the text black areas have some unwanted detail attributable to paper texture or print process, then lowering the black level can put this dark but visible detail below black making the text appear more cleaner.
Alternatively, if there is some black detail which is important, say the folds of a dark suit, the strands of black hair or the detail in a generally low-light image, then the raising of the Master Black level will provide visible detail in these dark areas.
Auto-Iris Detect
The auto-iris detect circuit controls the way the camera compensates for over-bright scenes. There are two aspects to it, namely control of the lens’ mechanical iris and of the electronic shutter (see Motion Control in the Systems section). The technique of controlling the exposure time is called Electronic Extended Iris (EEI).
Most imaging cameras have three modes for this circuit.
NORMAL : a mixture of peak and average
PEAK : iris is controlled by highlights
AVERAGE : iris is controlled by the average level
Peak would normally be used if flare from bright light sources was unwanted. This would enable one to read the text on an illuminated light bulb, for example. Average and normal would accept bright flare to avoid the blacking out of the majority scene.
Gain
For scenes of low light, amplification of the signal can restore the overall light level. As mentioned in Specifications and Units, this has many implications for sensitivity and signal to noise.
Most cameras offer the introduction of gain in discreet steps, e.g. +6dB, +9dB,+12dB and +18dB.
Other modes to be found in cameras are:
ALC :Automatic adjustment of the amount of gain required
ALC +EEI :Auto level adjustment from bright to dark scenes
Some imaging cameras have RS232 ports and control commands which can be used to achieve precise amounts of gain, for example 3.8 dB. This can be useful to introduce only the absolute minimum amount of gain required, thereby preserving the maximum S/N ratio.
Specialist Features
For an imaging camera to be truly versatile, it is essential to have a variety of functions allowing good imaging of difficult scenes. Many manufacturers have designed into their cameras a host of specialist modes.
Variable Scan
This allows the matching of the CCD exposure time to the frequency of a repeated image. Again it is best to illustrate by example: consider the case where a camera is required to film from a computer screen. These typically refresh at 72Hz, this means that every 14 ms a new frame is generated. If this is filmed with a 20 ms period, (50Hz), all of the first frame and some of the next will have been stored on the CCD, causing a narrow bright band where the overlap occurred. This band will scroll down the page, resulting in a poor and distracting image.
Variable scan will allow the CCD to “integrate” for shorter periods of time (as with electronic shutter). Thus, by selecting 1/72 as the time period, the CCD will store exactly one computer frame each time.
Slow Shutter
This is an extension of the above mode. It allows the CCD to be exposed for periods greater than one frame allowing the recovery of an image from a very dark scene. This dramatically improves the sensitivity of the camera without any reduction in the S/N ratio. In fact, the S/N ratio is improved by using Slow Shutter mode.
The slow shutter will allow integration in frame units from 1 to 200. This is particularly useful in the field of fluorescent microscopy where sub-lux light levels are often viewed. During the period of integration the video output is black and will appear stroboscopic. Obviously, the longer the integration period, the longer the period of black.
The best way to view this output is via a capture card that will allow a triggered input. Shortly before the camera outputs its accumulated image, it sends out a signal. This signal when fed into an appropriate capture card, will synchronize the card with the camera and refresh the screen only when a new image is received.
High Resolution Mode
This function was mentioned in the section on Resolution. Hi-Resolution it is an important feature, and is only really useful for still imaging. If the subject is not moving fast the camera should always be in this mode. For slow shutter, it is important to have hi-resolution on, otherwise the camera will only output one field not a full frame, this would reduce the vertical resolution by 50%.
This mode works by allowing the CCD to use the odd number rows of pixels for the odd field and the even rows for the even fields. Conventional cameras integrate twice per frame and use all of the CCD for each field. To compensate for the inevitable lost sensitivity, the camera integrates for longer for each field, this results in a noticeable picture lag should the subject move.
Stroboscopic illumination
High resolution mode is of considerable importance for stroboscopic illumination. Whenever the strobe’ fires, the light will fall in the integration period of each field. In normal mode, the light will only fall in the integration for the odd or the even field. This will result in a dramatic loss of vertical resolution.
If the strobe is asynchronous to the camera, and the image is to be captured on a capture card, the card will have to be synchronized with the strobe and accept odd/even or even/odd fields. If the card can only accept the conventional odd/even field order, then an occasional drop-out of a field will occur.
Random Trigger
Some applications require a camera’s vertical frame rate to be synchronized with an external event. This could be items on a fast moving conveyor belt, for example. By using a microswitch or line interrupter, a signal fed into the camera’s reset input, will reset the camera, to output a video frame at the right time, that is, when the object is in the middle of the frame. To help freeze the image, electronic shutter can be selected up to 1/10000th of a second. Linking this reset with the trigger of the capture card can produce stunning effects.