Glossary of Image Analysis Terms - A to F

Glossary of Image Analysis Terms (A to F)

A

Aberration – The failure of an optical lens to produce an exact point-to-point correspondence between the object and its resulting image. Various types are chromatic, spherical, coma, astigmatism and distortion.

Absorption – The loss of light of certain wavelengths as it passes through a material and is converted to heat or other forms of energy.

AC Coupled – A method of connecting a video signal to any circuit in a way that removes the potentially harmful DC offset. The DC offset is the overall voltage level that the video signal “rides” on.

Acquisition – The manner in which outside information is brought into an analysis system.

Active Illumination – Lighting a scene with a light source coordinated with the acquisition of an image. Strobed flash tubes, pulsed lasers and scanned LIDAR beams are examples.

Active Video – The part of the video waveform that is actually visible on the screen.

ActiveX® Control – Formerly known as OLE controls. A special form of component Automation Object that can be freely plugged into any OLE-enabled development tool, application or web browser.

A/D – See Analog-to-Digital Converter.

Algorithm –A set of well-defined rules or procedures for solving a problem or providing an output from a specific set of inputs.

Alpha Channel – A means by which two images can be blended, switched or combined in some way. Alpha numbers are stored in a separate plane of memory that sits alongside the frame buffer, one alpha number per pixel. The number stored determines what percentage of the first image is to be combined with what percentage of the second image.

Ambient Light – Light which is present in the environment of the imaging front end of a system and generated from outside sources. This light, unless used for actual scene illumination, will be treated as background noise by the imaging system.

Analog –A smooth, continuous voltage or current signal or function whose magnitude (value) is the information. From the word "analogous," meaning "similar to."

Analog-to-Digital Converter (A/D) –A device which converts an analog voltage or current signal to a discrete series of digitally encoded numbers (signal) for computer processing.

Aperture Delay – The time from the falling edge of the input clock of the A/D until the time the image is taken.

Aperture Jitter – Uncertainty in the aperture delay.

API – Application Programming Interface.

Area – Portion or area of the image to be analyzed. Area analysis measures the number of pixels which fall in a specified range of gray levels for the feature of interest.

Area Array Camera – A solid state imaging device with both rows and columns of pixels, forming an array which produces a 2D image.

Artifact – Blemishes, noise, snow, spots, or other physical distortions in an image. .

Aspect Ratio – The ratio of the width to the height of a frame of a video image. The U.S. television standard is 4:3 or 1.333

Astigmatism – A defect in a lens which causes blur or imperfect image results, since the rays from a given point fail to meet at the focal point.

Asynchronous – A camera characteristic which allows the return to top-of-frame to occur on demand, rather than synchronously following the 60 Hz power line scanning frequency.

Auto Focus – The ability of an imaging system to control the focus of the lens to obtain the sharpest image on the detector. Edge crispness is a typical control variable.

B

Background – The part of a scene behind the object to be imaged.

Backlighting – Placement of a light source behind an object so that a silhouette of that object is formed. It is used where outline information of the object and its features is important rather than surface features.

Back Porch – The area of the video waveform staked out by the rising edge of the horizontal sync pulse and right before the active video or before the terminating edge of the horizontal blanking.

Back Propagation – A training technique which adjusts the weights of the hidden and input layers of a neural net to force the correct decision for a given feature vector data input set.

Baffle – A type of shield that prohibits light from entering an optical system.

Bandpass Filter – An absorbing filter which allows a known range of wavelengths to pass, blocking those of lower or higher frequency.

Barrel Distortion – An optical imperfection which causes an image to bulge convexly on all sides similar to a barrel.

Beamsplitter – An optical device which divides one beam into two or more separate beams. A simple coated piece of glass in the optical path might reflect 60% of the light down onto the object, while allowing the other 40% to pass.

Binary – An image with pixel values either one or zero.

Binary Image – A black and white image represented as a single bit containing either zeros or ones, in which objects appear as silhouettes. The result of backlighting or thresholding.

Bit – An acronym for a BInary digiT. It is the smallest unit of information which can be represented. A bit may be in one of two states, on or off, represented by a zero or a one.

Bit Map – A representation of graphics or characters by individual pixels arranged in rows and columns. Black and white require one bit, while fancy high definition color up to 32.

Black Burst – The video waveform without the active video part. Black Level – This level represents the darkest an image can get. Defines what black is for a particular image system.

Blanking – The time during a raster scan retrace when the video signal is suppressed.

Blanking Level – That level of the video waveform defined by the system to be where blanking occurs. This could be the black level if a pedestal is used or below the black level if a pedestal is not used.

Blob – A single, connected region in a binary or grayscale image.

Blob Analysis – Identification of segmented objects in an image based on their geometric features (i.e. area, length, number of holes).

Blooming – An effect, sometimes caused when video becomes whiter-than-white, in which a thin line becomes larger and fuzzy on the screen.

Borescope – A device for internal inspection of difficult access locations such as pipes, engines, rifle barrels and pipes. Its long narrow tube contains a telescope system with a number of relay lenses. Light is provided via the optical path or fiber bundles. A 45 degree mirror at the end allows inspection of tube walls.

Boundary – The line formed by the joining of two image regions, each having a different light intensity. The edge of a region or object.

Bounding Box – The four coordinates which define a box around the object parallel to the major and minor axis.

Breezeway – That portion of the video waveform that sits between the rising edge of the horizontal sync and before the start of burst.

Brewster's Angle – The angle at which incident light, by reflecting at a boundary between two mediums of different refractive indices (i.e. air/glass or air/water), becomes plane polarized. For air/glass it is about 67.4 degrees.

Brightness – The total amount of light or incident illumination on a scene or object per unit area. Also called intensity.

Burst – See Chroma Burst.

Burst Gate – A signal that tells the system where the burst is located within the scan line.

Bus – Conductors used to interconnect individual circuitry within a computer. The conductors as a whole are called a bus.

Byte – Eight bits of digital information. A byte has values from 0 to 255, and is the unit most common to represent the gray scale value of one pixel.

C-mount – A threaded means of mounting a lens to a camera.

Calibration – The act of relating X and Y pixel spacing to known or predetermined pixels per unit length (i.e. inch, mm) factor. Often involves adjusting the image position in setup.

CCD (Charge Coupled Device) – A photo-sensitive image sensor implemented with large scale integration technology.

Centroid – The center of mass of an object having a constant density, or of an object having varying density, weighted by the gray scale value.

Character – A single letter, digit or punctuation symbol requiring one byte storage.

Chroma – See Chrominance.

Chroma Bandpass – In an NTSC or PAL video source the luma (black and white) and the chroma (color) information are combined together. In order to display an NTSC or PAL source, the luma and chroma must be split apart. The chroma bandpass filter removes the luma information from the chroma information, leaving the chroma information relatively intact.

Chroma Burst – The portion of the video waveform that helps the received determine what the colors are.

Chroma Demodulator – After the NTSC or PAL video source makes its way through the Y/C split, by either the chroma bandpass/trap method or the comb filter method, the color is decoded by the chroma demodulator.

Chroma Key – A method of combining two video images.

Chroma Trap – Similar to chroma bandpass, a method for separating the chroma information from the luma information, leaving the luma relatively intact.

Chromatic Dispersion – See Dispersion.

Chrominance – The color part of a video signal, relating to the hue and saturation but not the brightness or luminance of the signal. Also called Chroma.

Chrominance Filter – An electronic circuit, usually in the form of a notch filter, used to remove chrominance color information from a video signal.

CID – Charge Injection Device - A photo-sensitive image sensor implemented with large scale integration technology. Based on charge injection technology, a CID can be randomly addressed, non-destructively read, can be subscanned in a small region and is less susceptible to charge overflow from bright pixels to neighbors. The pixel structure is contiguous with maximum surface to capture incident light which is useful for sub-pixel measurement.

CIE – An acronym for a chromaticity coordinate system developed by the Commission Internationale de l'Eclairage, the international commission on illumination. In the CIE system, a plot of ratios (x, y and z) of the three standard primary colors (tristimulus values) to their sum. The most common diagram is the 2 dimensional CIE (x,y).

CIF (Common Interchange Format) – Developed to share images between computers.

Clamp – Also DC-Restoration. Restoring a DC component that was removed to make an AC coupled signal.

Classification – Assignment of image objects to one of two or more possible groups. Decisions are made by evaluating features either 1) structurally based on relationships or 2) statistically. For

example, 1) a penny is round, a certain diameter (+/- a tolerance) and has a histogram of a mean value; or 2) statistically, the object is measured a number of times, then the average and standard deviation are recorded. After training the features are weighted based on significance in object identification. For multiple features, absolute values are used.

Clipping Logic – A circuit used to prevent illegal conversion. After the conversion from one color space to another, a color space converter checks for illegal colors, and clipping logic is used to remove information until a legal color is represented.

Closing – A dilation followed by an erosion. A morphological operator useful to close holes and boundaries.

CMYK – A color space – Cyan, Magenta, Yellow, Black.

Coaxial Illumination – Front lighting with the illumination path running along the imaging optical axis and usually introduced with a 45 degree angle beam splitter.

Coherent Fiber Optics – A bundle of optical fibers with the input and output spatial x-y relationship maintained, resulting in near spatially correct image transmission.

Collimate – To produce light with parallel rays.

Collimated Lighting – Radiation from a given point with every light ray considered parallel. In actuality, even light from a very distant point source (i.e. a star) diverges somewhat. Note that all collimators have some aberrations.

Color – A visual object attribute which may be described by a "coordinate system" such as hue, saturation and intensity (HSI), CIE or LAB. Wavelengths in the visible part of the electromagnetic spectrum to which retinal rods respond.

Color Bar – A test pattern used to check whether a video system is calibrated correctly.

Color Burst – That portion of the video waveform that sits between the breezeway and the start of active video. It tells the color decoder in the receiver how to decode the color information contained in the next line of active video.

Color Decoder – The circuit in the video system that uses the chrominance portion of NTSC/PAL to derive the two color difference signals.

Color Difference – All of the color spaces used in imaging require three components (except for CMYK).

Color Encoder – Takes two color difference signals and combines them into the chroma signal. The opposite of the color decoder.

Color Key – See Chroma Key.

Color Killer – A circuit that shuts off the color decoder in a video system if the incoming video does not contain color difference signals.

Color Purity – Describes how close a color is to the theoretical.

Color Space – A two or three dimensional space used to represent an absolute color coordinate. RGB, HSI, LAB and CIE are all representations of color spaces.

Color Subcarrier – A clock signal used to run the color encoder or color decoder in an NTSC or PAL video system.

Color Temperature – A colorimetric concept related to the apparent visual color of a source, but not its actual temperature.

Colorimetry – Techniques used to measure color of an object or region and to define the results in a comparison or coordinate system.

Comb Filter – Used in place of a chroma bandpass and chroma trap combination – provides better video quality because it is more successful at separating the luma from the chroma.

Composite Sync – A signal consisting of horizontal sync pulses, vertical sync pulses, and equalizing pulses only, with a no-signal reference level.

Composite Video – A video stream which is produced by combining both a video and picture signal with horizontal and vertical sync and blanking signals.

Compression Ratio – A number which tells how much information has been compressed in an image.

Condenser Lens – Used to collect and redirect light for the purpose of illumination. Often used to collect light from a small source and project even light onto an object.

Connectivity Analysis – An SRI routine used to determine which pixels are interconnected and part of the same object or region. The results are used for blob analysis.

Contouring – An image artifact caused by not having enough bits to represent the image.

Contrast – The difference of light intensity between two adjacent regions in the image of an object. Often expressed as the difference between the lightest and darkest portion of an image. Contrast between a flaw or feature and its background is the goal of illumination.

Contrast Enhancement – Stretching of the gray level values between dark and light portions of an image to improve both visibility and feature detection.

Convolution – Superimposing an m x n operator (usually a 3x3 or 5x5 mask) over an area of the image, multiplying the points together, summing the results to replace the original pixel with the new value. This operation is often performed on the entire image to enhance edges, features, remove noise and other filtering operations.

Correlation – A mathematical measure of the similarity between images or areas within an image. Pattern matching or correlation of an X by Y array size template to the same size image, produces a scaler number, the percentage of match. Typically, the template is walked through a larger array to find the highest match.

Creepy-Crawlies – A specific image artifact resulting from the NTSC system. Also called zipper.

Cropping – A method of selecting a desired portion or region of interest out of an entire image.

Cross Section – A 3D profile of a slice of an object.

D

DAC – See Digital-to-Analog Converter.

Darkfield Illumination – Lighting of objects, surfaces or particles at very shallow or low angles, so that light does not directly enter the optics. Objects are bright with a dark background. This grazing illumination causes specular reflections from abrupt surface irregularities.

Data Reduction – The process of lowering the data content of a pixel or image such as thresholding or run length encoding.

DC Restoration – Restoring a DC component that was removed to make an AC coupled signal. Also see Clamp.

Decimation – A method of image scaling which involves removing unused pixels.

Decimation Filter – A filter designed to provide decimation without the artifacts associated with removing data.

Decision Tree – A structural classification technique based on relationships of feature measurements. Useful for differentiating a number of objects.

Dedicated System – Refers to a system which is configured for a specific application. Able to function when plugged in with no further development. Also called turnkey.

Demodulation – The technique used to separate the color components in NTSC or PAL systems.

Depth-of-Field – The range of an imaging system in which objects are in focus.

Depth Perception (3D) – Measurement of the third dimension of an object or scene.

Dichroic Filter – A filter used to transmit light based on its wavelength, rather than on its plane of vibration. Transmits one color, while reflecting a second when illuminated with white light. Often used in heads-up displays.

Differential Gain – When the magnitude of the chroma changes with respect to the luma when it should not, resulting in incorrect saturation.

Differential Phase – A change in the phase of the color burst for a change in the brightness of the picture.

Diffraction Pattern Sampling – Inspection by comparing portions of the interference pattern formed on a screen or special sensor from light waves diffracted by object edges.

Diffuse Reflection – Light which bounces off an object surface in many different directions. Light radiated from a matte surface is highly diffused.

Diffused Lighting – Scattered soft lighting from a wide variety of angles used to eliminate shadows and specular glints from profiled, highly reflective surfaces.

Digital Camera – The newest generation of video cameras transforms visual information into pixels, and then translates each pixel's level of light into a number in the camera.

Digital Clock Sync – A precise digital circuit that can quickly and accurately lock onto incoming video timing pulses, resulting in significantly lower jitter.

Digital Image – A video image converted into pixels. The numeric value of each pixel's value can be stored in a computer memory for subsequent processing and analysis.

Digital Imaging – Conversion of a video picture into pixels by means of an A/D converter where the level of each pixel can be stored in a computer.

Digital-to-Analog Converter (DAC) – A VLSI circuit used to convert digital computer processed images to analog for display on a monitor.

Digitization – Sampling and conversion of an incoming video or other analog signal into a digital value for subsequent storage and processing.

Dilation – A morphological operation which moves a probe or structuring element of a particular shape over the image, pixel by pixel. When an object boundary is contacted by the probe, a pixel is preserved in the output image. The effect is to "grow" the objects.

Dispersion – Separation of a beam of light into its wavelength components, each of which travel at slightly different speeds. Also called Chromatic Dispersion.

Dithering – Changing a range of gray to patterns of black and white, or using two or more colors to create the appearance of a larger section.

Dynamic Range – The measure of the range light sensitivity a sensor is able to reproduce, from the darkest to the brightest portion of a scene. Usually expressed in decibels.

E

Edge – A change in pixel values exceeding some threshold amount. Edges represent borders between regions on an object or in a scene.

Edge Detection – The ability to determine the true edge of an object.

Edge Enhancement – Image processing method to strengthen high-spatial frequencies in the image.

Edge Operator – Templates for finding edges in images.

Electrical Noise – Interference from various electrical devices which is present in the air as electromagnetic radiation or rides on the power lines and can introduce error into low voltage computations such as A/D conversion.

Electro-magnetic Spectrum – The total range of wavelengths, extending from the longest (audio) to the shortest (gamma rays) which can be physically generated. This entire spectrum is potentially useful for imaging, well beyond just the visible spectrum.

Encoder (Shaft or Position) – Provides rotation information for control of image acquisition, especially for moving web processes. Outputs either pulses for counting or BCD parallel with absolute position information.

Endoscope – A medical instrument used to view inside the human body. It may use borescope optics or coherent fibers to relay the image to the eye or camera. Illumination is provided by a non-coherent bundle of optical fibers.

Erosion – The converse of the morphology dilation operator. A morphological operation which moves a probe or structuring element of a particular shape over the image, pixel by pixel. When the probe fits inside an object boundary, a pixel is preserved in the output image. The effect is to "shrink or erode" objects as they appear in the output image. Any shape smaller than the probe

(i.e. noise) disappears.

Extension Tube – A cylindrical threaded tube used to change the magnification, effective focal length and field of view of a lens when inserted between the lens and imaging sensor.

F

F-number or f-stop – The ratio of the focal length to the lens aperture. The smaller the f- number, the larger the lens diameter and brighter the image and narrower the depth-of-field.

Fade – A method of switching from one video source to another video source.

Fast Fourier Transform – Produces a new image which represents the frequency domain content of the spatial or time domain image information. Data is represented as a series of sinusoidal waves.

Feature Extraction – Determining image features by applying feature detectors to distinguish or segment them from the background.

Feature Vectors – A set of features of an object (such as area, number of holes, etc) that can be used for its identification or inspection.

Features – Simple image data attributes such as pixel amplitudes, edge point locations and textural descriptors, center of mass, number of holes in an object with distinctive characteristics defined by boundaries or regions.

Fiber Optics – Light source or optical image delivery via a long, flexible fiber(s) of transparent material, usually bundled together. Light is transmitted via internal reflection inside each fiber. Coherent fiber optics are spatially organized so images can be relayed.

Fiberscope – An optical instrument similar to a borescope, but uses a flexible, coherent fiber or bundle (usually silicon), an objective lens and an eyepiece or camera.

Fiducial – A line, mark or shape used as a standard of reference for measurement or location.

Field – The set of either the even or odd lines in a video image. The odd plus the even field comprise one video frame. A field is scanned every 1/60th of a second.

Field-of-View (FOV) – The 2D area which can be seen through the optical imaging system.

Filter – A device or process that selectively transmits frequencies. In optics, the material either reflects or absorbs certain wavelengths of light, while passing others. Filters can be used to sharpen or smooth image details, or to extract edges from objects.

Filtering – The use of an optical filter for picture or color enhancement in front of the camera lens or light source. Also analog or digital image processing operations to enhance or modify an image. May be linear or non-linear.

Firmware – Software hard coded in non-volatile memory (ROM), usually to increase speed.

Fixture – A device to hold and locate a workpiece during processing or inspection operations.

Flicker – Occurs when the frame rate of the video is too low and the human eye can perceive the individual frames.

Fluorescence – The emission of light or other electromagnetic radiation at longer wavelengths by matter as a result of absorption of a shorter wavelength. The emission lasts only as long as the stimulating irradiation is present.

Focal Length – The distance from a lens' principal point to the corresponding focal point on the object.

Focal Plane – Usually found at the image sensor, it is a plane perpendicular to the lens axis at the point of focus.

Focus – The point at which rays of light converge for any given point on the object in the image. Also called the focal point.

Focus Following – A ranging and tracking technique that uses image processing to measure object range based on best focus.

Fourier Domain Inspection – Evaluation of the Fourier transform (frequency information) of a 2D spatial image for features of interest.

Frame – The total area scanned in an image sensor while the video signal is not blanked. Essentially one picture out of a video stream. In interlaced scanning, two fields comprise one frame. Frame rate is typically 30 Hz.

Frame Buffer – Image memory in a frame grabber.

Frame Grabber – A device that interfaces with a camera and, on command, samples the video, converts the sample to a digital value and stores that number in a computer's memory.

Frame Rate – The frequency at which an image is completely updated on the display monitor.

Frame Transfer CCD (Charge Coupled Device) – The entire image is transferred from the sensing area to a storage area on a chip. Data (charge) is read out from the storage area in a full frame mode. This workhorse of the industry is also capable of non-RS-170 operation.

Front End System – The object, illumination, optics and imager blocks of an imaging system. Includes all components useful to acquire a good image for subsequent processing.

Front Lighting – The use of illumination on the camera side of an object so that surface features can be observed.

Front Porch – The area of the video waveform that sits between the start of horizontal blank and the falling edge (start) of horizontal sync.