Glossary

NI Vision 2015 Concepts Help

Edition Date: June 2015

Part Number: 372916T-01

»View Product Info
Download Help (Windows Only)

Numbers/Symbols

1DOne-dimensional.
2DTwo-dimensional.
3DThree-dimensional.

A

acceptance levelValue that indicates how closely an object must match a trained character to be recognized. A high acceptance level value indicates that the object and trained character must be closely matched for the object to be recognized. (OCR)
AIPDThe National Instruments internal image file format used for saving complex images and calibration information associated with an image. AIPD images have the file extension APD.
alignmentThe process by which a machine vision application determines the location, orientation, and scale of a part being inspected.
alpha channelChannel used to code extra information, such as Gamma correction, about an image.
angle distanceIn contour analysis, angle distance is the amount, in degrees, that one curve endpoint must rotate for two curves to be parallel at the endpoints.
areaA rectangular portion of an acquisition window or frame that is controlled and defined by software.
area thresholdDetects objects based on their size.
arithmetic operatorsThe image operations multiply, divide, add, subtract, and remainder.
arrayAn ordered, indexed set of data elements of the same type.
artifactExtraneous pixels in the region of interest. (OCR)
aspect ratioHeight/width ratio of a character. (OCR)
auto-median functionA function that uses dual combinations of opening and closing operations to smooth the boundaries of objects.
AutoSplitWorks in conjunction with the maximum character bounding rectangle width, and uses an algorithm to analyze the right side of a character bounding rectangle. AutoSplit then determines the rightmost vertical line in the object that contains the fewest number of pixels and moves the rightmost edge of the character bounding box to that location. (OCR)

B

bBit. One binary digit, either 0 or 1.
BByte. Eight related bits of data, an 8-bit binary number. Also denotes the amount of memory required to store one byte of data.
barycenterThe value representing the centroid of the range of an image’s grayscale values in the image histogram.
binary imageAn image containing objects usually represented with a pixel intensity of 1 (or 255) and the background of 0.
binary morphologyFunctions that perform morphological operations on a binary image.
binary thresholdThe separation of an image into objects of interest (assigned pixel values of 1) and background (assigned pixel values of 0) based on the intensities of the image pixels.
bit depthThe number of bits (n) used to encode the value of a pixel. For a given n, a pixel can take 2n different values. For example, if n equals eight bits, a pixel can take 256 different values ranging from 0 to 255. If n equals 16 bits, a pixel can take 65,536 different values ranging from 0 to 65,535 or –32,768 to 32,767.
black reference levelThe level that represents the darkest value an image can have. See also white reference level.
blurringReduces the amount of detail in an image. Blurring can occur when the camera lens is out of focus or when an object moves rapidly in the field of view. You can blur an image intentionally by applying a lowpass frequency filter.
BMPBitmap. Image file format commonly used for 8-bit and color images (extension .bmp).
border functionRemoves objects (or particles) in a binary image that touch the image border.
brightness(1) A constant added to the red, green, and blue components of a color pixel during the color decoding process; (2) The perception by which white objects are distinguished from gray and light objects from dark objects.
bufferTemporary storage in memory for acquired data.

C

caliperA function that calculates distances, angles, circular fits, and the center of mass based on positions given by edge detection, particle analysis, centroid, and search functions.
cellA single module that encodes one bit of data in a 2D code.
center of massThe point on an object where all the mass of the object could be concentrated without changing the first moment of the object about any axis.
characterRecognized group of foreground elements. (OCR)
character bounding rectangleSmallest rectangle that completely encloses a character. (OCR)
character classIn OCR, a group of characters that have been trained with the same character value.
character recognition Ability of a machine to read human-readable text. (OCR)
character segmentationApplication of several parameters, such as thresholding, character size, and element spacing, that isolates a character in a region of interest. (OCR)
character setSet of trained characters and/or patterns. (OCR)
character set fileFile that contains a character set. (OCR)
character sizeNumber of pixels that make up a character. (OCR)
character spacingIn OCR, the horizontal distance between the right edge of one character bounding rectangle and the left edge of the next character bounding rectangle.
character valueString that describes a character. For example, you might assign the character value "A" to a group of elements that resembles the letter A. (OCR)
chromaThe color information in a video signal.
chromaticityThe combination of hue and saturation. The relationship between chromaticity and brightness characterizes a color.
chrominanceSee chroma.
circle functionDetects circular objects in a binary image.
classA category representing a collection of similar samples.
classificationAn operation that assigns samples to classes based on predefined features.
classification accuracyThe probability that a sample is classified into the class to which it belongs.
classification confidenceThe degree of certainty that a sample or character is assigned to one class instead of other classes. See also class and sample.
classification predictive valueThe probability that a sample classified into a given class belongs to that class.
classification scoreSee classification confidence.
classifierA function or VI that assigns a sample to a class.
closingA dilation followed by an erosion. A closing fills small holes in objects and smooths the boundaries of objects.
clusteringTechnique where the image is sorted within a discrete number of classes corresponding to the number of phases perceived in an image. The gray values are determined and a barycenter is determined for each class. This process is repeated until a value is obtained that represents the center of mass for each phase or class.
CLUTSee color look-up table.
code wordNumeric value of the printed bar/space pattern in a 1D or 2D barcode.
color imagesImages containing color information, usually encoded in the RGB form.
color spaceThe mathematical representation for a color. For example, color can be described in terms of red, green, and blue; hue, saturation, and luminance; or hue, saturation, and intensity.
complex imageStores information obtained from the FFT of an image. The complex numbers that compose the FFT plane are encoded in 64-bit floating-point values: 32 bits for the real part and 32 bits for the imaginary part.
color look-up table (CLUT)Table for converting the value of a pixel in an image into a red, green, and blue (RGB) intensity.
connectivityDefines which of the surrounding pixels of a given pixel constitute its neighborhood. See also connectivity-4 and connectivity-8.
connectivity distanceIn contour analysis, connectivity distance is the minimum distance required to connect two endpoints when a line extended from one endpoint passes within the minimum distance of the second endpoint.
connectivity-4Connectivity where only pixels adjacent in the horizontal and vertical directions are considered neighbors. Two pixels are considered as part of a same object if they are horizontally or vertically adjacent. They are considered as part of two different objects if they are diagonally adjacent.
connectivity-8Connectivity where all adjacent pixels are considered neighbors. Two pixels are considered as part of a same object if they are horizontally, vertically, or diagonally adjacent.
convex hullThe smallest convex polygon that can encapsulate a particle.
convex hull functionComputes the convex hull of objects in a binary image.
convolutionSee linear filter.
convolution kernelSimple 3 × 3, 5 × 5, or 7 × 7 matrices (or templates) used to represent the filter in the filtering process. The contents of these kernels are a discrete two-dimensional representation of the impulse response of the filter that they represent.
coordinate systemA reference location (origin) and angle in an image that regions of interest can relate to when positional and angular adjustments of the region of interest are necessary. A coordinate system is depicted by two lines representing the orientation and direction of its two axes.
cross correlationThe most common way to perform pattern matching.
curvea set of edge points that are connected to form a continuous contour.
curve extractionThe process of finding curves, or connected edge points, in a grayscale image. Curves usually represent the boundaries of objects in the image.

D

Danielsson functionSimilar to the distance functions, but with more accurate results.
dBDecibel. A unit for expressing a logarithmic measure of the ratio of two signal levels: dB = 20log10 V1/V2, for signals in volts.
default settingA default parameter value recorded in the driver. In many cases, the default input of a control is a certain value (often 0).
defect imageAn image that shows all the defects found during golden template comparison.
definitionThe number of values a pixel can take on, which is the number of colors or shades that you can see in the image.
dendriteBranches of the skeleton of an object.
densitometryThe determination of optical or photographic density.
density functionFor each gray level in a linear histogram, it gives the number of pixels in the image that have the same gray level.
differentiation filterExtracts the contours (edge detection) in gray level.
digital imageAn image f (x, y) that has been converted into a discrete number of pixels. Both spatial coordinates and brightness are specified.
dilationIncreases the size of an object along its boundary and removes tiny holes in the object.
distanceDetermination of the physical dimensions of a pixel by defining the physical dimensions of a line in the image.
distance calibrationIn contour analysis, distance is the euclidean distance, in pixels, between two endpoints.
distance functionAssigns to each pixel in an object a gray-level value equal to its shortest Euclidean distance from the border of the object.
distance metricA metric that computes the distance between features in a classification application.
distribution functionFor each gray level in a linear histogram, the function gives the number of pixels in the image that are less than or equal to that gray level.
dot-matrix characterCharacter comprised of a series of small elements. See segmented character. (OCR)
driverSoftware that controls a specific hardware device, such as a data acquisition board.
dynamic rangeThe ratio of the largest signal level a circuit can handle to the smallest signal level it can handle (usually taken to be the noise level), normally expressed in decibels.

E

edgeDefined by a sharp change (transition) in the pixel intensities in an image or along an array of pixels.
edge contrastThe difference between the average pixel intensity before the edge and the average pixel intensity after the edge.
edge detectionAny of several techniques that identify the edges of objects in an image.
edge hysteresisThe difference in threshold level between a rising and a falling edge.
edge steepnessThe number of pixels that corresponds to the slope or transition area of an edge.
elementConnected group of foreground pixels. Adjacent elements form a character. See object. (OCR)
element spacingAmount of space, in pixels, between elements. See horizontal element spacing and vertical element spacing. (OCR)
energy centerThe center of mass of a grayscale image. See also center of mass.
entropyMethod based on a classical image analysis technique that is best suited for detecting objects present in minuscule proportions on the image. For example, this function would be suitable for default detection.
equalize functionSee histogram equalization.
erasureMissing or undecodable codeword at a known position in a 2D barcode.
erosionReduces the size of an object along its boundary and eliminates isolated points in the image.
Euclidean distanceThe shortest distance between two points in a Cartesian system.
exponential and gamma correctionsExpand the high gray-level information in an image while suppressing low gray-level information.
exponential functionDecreases the brightness and increases the contrast in bright regions of an image, and decreases contrast in dark regions.

F

featureA measurement from or attribute of a sample.
feature extractionAn operation that computes features of a sample.
feature vectorA 1D array in which each element represents a different feature of a sample.
FFTFast Fourier Transform. A method used to compute the Fourier transform of an image.
fiducialA reference pattern on a part that helps a machine vision application find the part location and orientation in an image. An alignment mark.
finder patternA specific pattern or sequence used by 2D codes to identify the orientation of data in the 2D code.
focal pointThe pixel location in a pattern matching template whose coordinates are returned as the location of a match in the inspection image.
Fourier spectrumThe magnitude information of the Fourier transform of an image.
Fourier transformTransforms an image from the spatial domain to the frequency domain.
frequency filtersCounterparts of spatial filters in the frequency domain. For images, frequency information is in the form of spatial frequency.
functionSet of software instructions executed by a single line of code that may have input and/or output parameters and returns a value when executed.

G

gammaThe nonlinear change in the difference between the video signal’s brightness level and the voltage level needed to produce that brightness.
gaugingMeasurement of an object or distances between objects.
Gaussian filterA filter similar to the smoothing filter, but using a Gaussian kernel in the filter operation. The blurring in a Gaussian filter is more gentle than a smoothing filter.
Gaussian filterA filter similar to the smoothing filter, but using a Gaussian kernel in the filter operation. The blurring in a Gaussian filter is more gentle than a smoothing filter.
geodesicRelating to or involving the geometry of curved surfaces.
geometric featuresInformation extracted from a grayscale template that are used to locate the template in the target image. Geometric features in an image range from low-level features such as edges or curves detected in the image to higher-level features such as the geometric shapes made by the curves in the image.
geometric matchingA technique used to locate quickly a grayscale template that is characterized by distinct geometric or shape information within a grayscale image.
golden templateAn image containing an ideal representation of an object under inspection.
gradient convolution filterSee gradient filter.
gradient distanceIn contour analysis, gradient distance is the difference between the gradient angles of two endpoints.
gradient filterExtracts the contours (edge detection) in gray-level values. Gradient filters include the Prewitt and Sobel filters.
gray levelThe brightness of a pixel in an image.
gray-level co-occurrence matrixA two-dimensional array of correlated pixel values in an image. The array values indicate the number of times that every pixel value in the image occurs at a given distance vector from any other pixel value in the image.
gray-level dilationIncreases the brightness of pixels in an image that are surrounded by other pixels with a higher intensity.
gray-level erosionReduces the brightness of pixels in an image that are surrounded by other pixels with a lower intensity.
grayscale imageAn image with monochrome information.
grayscale morphology Functions that perform morphological operations on a grayscale image.

H

Haralick featuresA set of second-order statistics derived from the gray-level co-occurrence matrix of an image.
highpass attenuationApplies a linear attenuation to the frequencies in an image, with no attenuation at the highest frequency and full attenuation at the lowest frequency.
highpass FFT filterRemoves or attenuates low frequencies present in the FFT domain of an image.
highpass filterEmphasizes the intensity variations in an image, detects edges (or object boundaries), and enhances fine details in an image.
highpass frequency filterAttenuates or removes (truncates) low frequencies present in the frequency domain of the image. A highpass frequency filter suppresses information related to slow variations of light intensities in the spatial image.
highpass truncationRemoves all frequency information below a certain frequency.
histogramIndicates the quantitative distribution of the pixels of an image per gray-level value.
histogram equalizationTransforms the gray-level values of the pixels of an image to occupy the entire range (0 to 255 in an 8-bit image) of the histogram, increasing the contrast of the image.
histogram inversionFinds the photometric negative of an image. The histogram of a reversed image is equal to the original histogram flipped horizontally around the center of the histogram.
hit-miss functionLocates objects in the image similar to the pattern defined in the structuring element.
hole filling functionFills all holes in objects that are present in a binary image.
horizontal element spacingSpace, in pixels, between horizontally adjacent elements. (OCR)
HSIColor encoding scheme in Hue, Saturation, and Intensity.
HSLColor encoding scheme in Hue, Saturation, and Luminance.
HSVColor encoding scheme in Hue, Saturation, and Value.
hueRepresents the dominant color of a pixel. The hue function is a continuous function that covers all the possible colors generated using the R, G, and B primaries. See also RGB.
hyperplaneIn SVM classification, a separating surface at the maximum possible distance from the support vectors in either of two classes.

I

identification confidenceThe degree of similarity between a sample and members of the class to which the sample is assigned.
imageLight intensity as a function of the spatial coordinates f (x, y) where x and y denote spatial coordinates and the value f at any point (x, y) is the light intensity at that point.
image borderA user-defined region of pixels surrounding an image. Functions that process pixels based on the value of pixel neighbors require image borders.
image browserAn image that contains thumbnails of images to analyze or process in a vision application.
image bufferA memory location used to store images.
image definitionSee pixel depth.
image display environment A window or control that displays an image.
image enhancementThe process of improving the quality of an image that you acquire from a sensor in terms of signal-to-noise ratio, image contrast, edge definition, and so on.
image fileA file containing pixel data and additional information about the image.
image formatDefines how an image is stored in a file. Usually composed of a header followed by the pixel data.
image maskA binary image that isolates parts of a source image for further processing. A pixel in the source image is processed if its corresponding mask pixel has a non-zero value. A source pixel whose corresponding mask pixel has a value of 0 is left unchanged.
image paletteThe gradation of colors used to display an image on screen, usually defined by a color look-up table.
image processingEncompasses various processes and analysis functions which you can apply to an image.
image segmentationThe process of separating objects from the background and each other so that each object can be identified and characterized.
image sourceOriginal input image. If you start processing an image and want to revert back to the original image, select Source Image from the list box. When you click on Source Image, the image reverts to the original image, and any processing done on the image prior to selecting Image Source is cancelled.
image visualizationThe presentation (display) of an image (image data) to the user.
imagingAny process of acquiring and displaying images and analyzing image data.
inner gradientFinds the inner boundary of objects.
inspectionA vision inspection application created in and ran from Vision Builder AI that can perform inline or offline visual inspections.
inspection functionsDetects specific features in an image. The features detected include edges, peaks, and rotational shifts.
intensityThe gray-level value of a pixel in a grayscale image.
intensity calibrationAssigning user-defined quantities, such as optical densities or concentrations, to the gray-level values in an image.
intensity profileThe gray-level distribution of the pixels along an ROI in an image.
intensity rangeDefines the range of gray-level values in an object of an image.
intensity thresholdCharacterizes an object based on the range of gray-level values in the object. If the intensity range of the object falls within the user specified range, it is considered an object; otherwise it is considered part of the background.
interpolationIs the technique used to find values in between known values when resampling an image or array of pixels.
invariant featureA feature vector that is invariant to changes in the scale, rotation, and mirror symmetry of samples.
IREA relative unit of measure (named for the Institute of Radio Engineers). 0 IRE corresponds to the blanking level of a video signal, 100 IRE to the white level. Notice that for CCIR/PAL video, the black level is equal to the blanking level or 0 IRE, while for RS-170/NTSC video, the black level is at 7.5 IRE.

J

JPEGJoint Photographic Experts Group. Image file format for storing 8-bit and color images with lossy compression (extension .jpg).

K

kernelStructure that represents a pixel and its relationship to its neighbors. The relationship is specified by weighted coefficients of each neighbor.

L

labelingThe process by which each object in a binary image is assigned a unique value. This process is useful for identifying the number of objects in the image and giving each object a unique identity.
LabVIEWLaboratory Virtual Instrument Engineering Workbench. Program development environment application based on the programming language G used commonly for test and measurement applications.
Laplacian filterExtracts the contours of objects in the image by highlighting the variation of light intensity surrounding a pixel.
line gaugeMeasures the distance between selected edges with high-precision subpixel accuracy along a line in an image. For example, this function can be used to measure distances between points and edges. This function also can step and repeat its measurements across the image.
line profileRepresents the gray-level distribution along a line of pixels in an image.
linear filterA special algorithm that calculates the value of a pixel based on its own pixel value as well as the pixel values of its neighbors. The sum of this calculation is divided by the sum of the elements in the matrix to obtain a new pixel value.
local thresholdA method of image segmentation that categorizes a pixel as part of a particle or the background based on the intensity statistics of the particle’s neighboring pixels.
logarithmic and inverse gamma correctionsExpand low gray-level information in an image while compressing information from the high gray-level ranges.
logarithmic functionIncreases the brightness and contrast in dark regions of an image, and decreases the contrast in bright regions of the image.
logic operatorsThe image operations AND, NAND, OR, XOR, NOR, difference, mask, mean, max, and min.
look-up tableTable containing values used to transform the gray-level values of an image. For each gray-level value in the image, the corresponding new value is obtained from the look-up table and applied to the image.
lossless compressionCompression in which the decompressed image is identical to the original image.
lossy compressionCompression in which the decompressed image is visually similar but not identical to the original image.
lowpass attenuationApplies a linear attenuation to the frequencies in an image, with no attenuation at the lowest frequency and full attenuation at the highest frequency.
lowpass FFT filterRemoves or attenuates high frequencies present in the FFT domain of an image.
lowpass filterAttenuates intensity variations in an image. You can use these filters to smooth an image by eliminating fine details and blurring edges.
lowpass frequency filterAttenuates high frequencies present in the frequency domain of the image. A lowpass frequency filter suppresses information related to fast variations of light intensities in the spatial image.
lowpass truncationRemoves all frequency information above a certain frequency.
L-skeleton functionUses an L-shaped structuring element in the Skeleton function.
lumaThe brightness information in the video picture. The luma signal amplitude varies in proportion to the brightness of the video signal and corresponds exactly to the monochrome picture.
luminanceSee luma.
LUTSee look-up table.

M

machine visionAn automated application that performs a set of visual inspection tasks.
marginIn SVM classification, the distance between the support vectors of a class and the hyperplane.
mask FFT filterRemoves frequencies contained in a mask (range) specified by the user.
match scoreA number ranging from 0 to 1000 that indicates how closely an area of an acquired image matches the template image. A match score of 1000 indicates a perfect match. A match score of 0 indicates no match.
median filterA lowpass filter that assigns to each pixel the median value of its neighbors. This filter effectively removes isolated pixels without blurring the contours of objects.
metricTechnique used in situations similar to interclass variance. For each threshold, a value is calculated that is determined by the surfaces representing the initial grayscale. The optimal threshold corresponds to the smallest value.
module
  1. A single cell that encodes one bit of data in a Data Matrix barcode.
  2. The narrowest discernible element in a PDF417 barcode.
momentsTechnique best suited for images that have poor contrast (an overexposed image is better processed than an underexposed image). The moments method is based on the hypothesis that the observed image is a blurred version of the theoretically binary original. The blurring that is produced from the acquisition process (electronic noise or slight defocalization) is treated as if the statistical moments (average and variance) were the same for both the blurred image and the original image. This function recalculates a theoretical binary image.
morphological transformationsExtract and alter the structure of objects in an image. You can use these transformations for expanding (dilating) or reducing (eroding) objects, filling holes, closing inclusions, or smoothing borders. They mainly are used to delineate objects and prepare them for quantitative inspection analysis.
M-skeletonUses an M-shaped structuring element in the skeleton function.
multiple template matchingThe technique used to simultaneously locate multiple grayscale templates within a grayscale image.

N

neighborA pixel whose value affects the value of a nearby pixel when an image is processed. The neighbors of a pixel are usually defined by a kernel or structuring element.
neighborhood operationsOperations on a point in an image that take into consideration the values of the pixels neighboring that point.
nonlinear filterReplaces each pixel value with a nonlinear function of its surrounding pixels.
nonlinear gradient filterA highpass edge-extraction filter that favors vertical edges.
nonlinear Prewitt filterA highpass edge-extraction filter that favors horizontal and vertical edges in an image.
nonlinear Sobel filterA highpass edge-extraction filter that favors horizontal and vertical edges in an image.
number of planes (in an image)The number of arrays of pixels that compose the image. A gray-level or pseudo-color image is composed of one plane, while an RGB image is composed of three planes (one for the red component, one for the blue, and one for the green).
Nth order filterFilters an image using a nonlinear filter. This filter orders (or classifies) the pixel values surrounding the pixel being processed. The pixel being processed is set to the Nth pixel value, where N is the order of the filter.

O

object A group of elements that satisfy element spacing requirements. An object is an unrecognized character. See also particle. (OCR)
occlusion invariant matching A geometric matching technique in which the reference pattern can be partially obscured in the target image.
OCROptical character recognition. The process of analyzing an image to detect and recognize characters/text in the image.
OCR SessionCharacter set and parameter settings that define an instance of OCR.
OCVOptical character verification. A machine vision application that inspects the quality of printed characters.
offsetThe coordinate position in an image where you want to place the origin of another image. Setting an offset is useful when performing mask operations.
openingAn erosion followed by a dilation. An opening removes small objects and smooths boundaries of objects in the image.
operatorsAllow masking, combination, and comparison of images. You can use arithmetic and logic operators in NI Vision.
optical representationContains the low-frequency information at the center and the high-frequency information at the corners of an FFT-transformed image.
outer gradientFinds the outer boundary of objects.
overlayRegions of interest, text, and bitmaps that you can place on top of a displayed image to annotate it without modifying it.

P

paletteThe gradation of colors used to display an image on screen, usually defined by a color look-up table.
particleConnected region or grouping of pixels in an image in which all pixels have the same intensity level.
article analysisA series of processing operations and analysis functions that produce information about the particles in an image.
patternCharacter for which the character value requires more than one byte. (OCR)
pattern matchingThe technique used to quickly locate a grayscale template within a grayscale image.
picture aspect ratioThe ratio of the active pixel region to the active line region. For standard video signals like RS-170 or CCIR, the full-size picture aspect ratio normally is 4/3 (1.33).
picture elementAn element of a digital image. Also called pixel.
pivot pixelThe pixel at the center of an inspection window.
pixelPicture element. The smallest division that makes up a digital image. For measurement tasks, optimum pixel dimensions are square (aspect ratio of 1:1, or the width equal to the height).
pixel aspect ratioThe ratio between the physical horizontal and vertical sizes of the region covered by the pixel. An acquired pixel should optimally be square, thus the optimal value is 1.0, but typically it falls between 0.95 and 1.05, depending on camera quality.
pixel calibrationDirectly calibrating the physical dimensions of a pixel in an image.
pixel depthThe number of bits (n) used to code the intensity of a pixel. For a given n, a pixel can take 2n different values. For example, if n equals 8-bits, a pixel can take 256 different values ranging from 0 to 255. If n equals 16 bits, a pixel can take 65,536 different values ranging from 0 to 65,535 or –32,768 to 32,767.
pixel frameDescribes the neighborhood. For a square pixel frame, each pixel is surrounded by eight neighbors. The vertical and horizontal neighbors have a distance d from the pixel. Diagonal pixels have a slightly greater distance because they are farther away from the central pixel. For a hexagonal pixel frame, each pixel is surrounded by six neighbors. Each neighbor is at an equal distance from the central pixel.
PNGPortable Network Graphic. Image file format for storing 8-bit, 16-bit, and color images with lossless compression (extension .png).
Power 1/Y functionSimilar to a logarithmic function but with a weaker effect.
Power Y functionSee exponential function.
Prewitt filterExtracts the contours (edge detection) in gray-level values using a 3 × 3 filter kernel.
probability functionDefines the probability that a pixel in an image has a certain gray-level value.
proper-closingA finite combination of successive closing and opening operations that you can use to fill small holes and smooth the boundaries of objects.
proper-openingA finite combination of successive opening and closing operations that you can use to remove small particles and smooth the boundaries of objects.
pyramidal matchingA technique used to increase the speed of a pattern matching algorithm by matching subsampled versions of the image and the reference pattern.

Q

quantitative analysisObtaining various measurements of objects in an image.
quiet zoneThe area containing no data that is required to surround a 2D barcode. This area is measured in module widths.

R

read resolutionLevel of character criteria OCR uses to determine if an object matches a trained character.
read strategyMethod by which you determine how stringently OCR analyzes objects to determine if they match trained characters.
readingIn OCR, the process of segmenting each object in an image and comparing it to trained characters to determine if there is a match. See also training and verifying.
real timeA property of an event or system in which data is processed as it is acquired instead of being accumulated and processed at a later time.
reference characterIn OCR, the character in a character class that is designated as the best representative of the character value for which all the characters in the class were trained.
reflectance valueA value used to determine the symbol contrast of a Data Matrix code in an image. The reflectance value of a pixel equals the pixel intensity divided by the maximum possible value for the given image type. (For example, 255 is the maximum possible pixel value for an 8-bit image.)
region of interest (ROI)An area of the image that is graphically selected from a window displaying the image. This area can be used to focus further processing.
relative accuracyA measure in LSB of the accuracy of an ADC; it includes all nonlinearity and quantization errors but does not include offset and gain errors of the circuitry feeding the ADC.
resolutionThe number of rows and columns of pixels. An image composed of m rows and n columns has a resolution of m´n. This image has n pixels along its horizontal axis and m pixels along its vertical axis.
reverse functionInverts the pixel values in an image, producing a photometric negative of the image.
RGBColor image encoding using red, green, and blue colors.
RGB U64A color encoding scheme using red, green, and blue (RGB) color information where each pixel in the color image is encoded using 64 bits:16 bits for red, 16 bits for green, 16 bits for blue, and 16 bits for the alpha value (unused).
Roberts filterExtracts the contours (edge detection) in gray level, favoring diagonal edges.
ROISee region of interest.
rotational shiftThe amount by which one image is rotated with respect to a reference image. This rotation is computed with respect to the center of the image.
rotation-invariant matching A pattern matching technique in which the reference pattern can be located at any orientation in the test image as well as rotated at any degree.

S

sampleAn object in an image that you want to classify.
saturationThe amount of white added to a pure color. Saturation relates to the richness of a color. A saturation of zero corresponds to a pure color with no white added. Pink is a red with low saturation.
scale-invariant matching A pattern matching technique in which the reference pattern can be any size in the test image.
seed pointA point on a curve from which tracing begins.
segmentation functionFully partitions a labeled binary image into non-overlapping segments, with each segment containing a unique object.
segmented characterCharacter that OCR isolates according to specific parameters, such as thresholding, character size, and so on.
separation functionSeparates objects that touch each other by narrow isthmuses.
shape descriptorA feature vector that describes the shape of a sample.
shape detectionDetects rectangles, lines, ellipses, and circles within images.
shape matchingFinds objects in an image whose shape matches the shape of the object specified by a template. The matching process is invariant to rotation and can be set to be invariant to the scale of the objects.
shift-invariant matching A pattern matching technique in which the reference pattern can be located anywhere in the test image but cannot be rotated or scaled.
Sigma filterA highpass filter that outlines edges.
skeleton functionApplies a succession of thinning operations to an object until its width becomes one pixel.
skiz functionObtains lines in an image that separate each object from the others and are equidistant from the objects that they separate.
smoothing filterBlurs an image by attenuating variations of light intensity in the neighborhood of a pixel.
Sobel filterExtracts the contours (edge detection) in gray-level values using a 3 × 3 filter kernel.
spatial calibrationAssigning physical dimensions to the area of a pixel in an image.
spatial filtersAlter the intensity of a pixel with respect to variations in intensities of its neighboring pixels. You can use these filters for edge detection, image enhancement, noise reduction, smoothing, and so forth.
spatial resolutionThe number of pixels in an image, in terms of the number of rows and columns in the image.
Square functionSee exponential function.
Square Root functionSee logarithmic function.
standard representationContains the low-frequency information at the corners and high-frequency information at the center of an FFT-transformed image.
stroke characterCharacter that consists of continuous elements in which breaks are caused only by imperfections in the image. (OCR)
structuring elementA binary mask used in most morphological operations. A structuring element is used to determine which neighboring pixels contribute in the operation and to what degree.
sub-pixel analysisUsed to find the location of an edge or template in terms of fractions of a pixel.
substitution characterCharacter that represents unrecognized characters. Typically, the substitution character is a question mark (?). (OCR)
substitution errorErroneously decoded codeword at an unknown position in a 2D barcode.
supervised learning methodA machine learning technique that requires labelled training data.
support vectorA sample in one class that is closest to another class.
support vector machine (SVM)A supervised learning method used for classification. SVM classifiers identify a hyperplane located at the maximum possible distance from the nearest support vector in either of two classes. SVM classifiers have very good generalization capabilities and perform well in high dimensional feature spaces.

T

templatePattern that you are trying to match in an image using the Match Pattern step. A template can be a region selected from an image or it can be an entire image.
thickeningAlters the shape of objects by adding parts to the object that match the pattern specified in the structuring element.
thinningAlters the shape of objects by eliminating parts of the object that match the pattern specified in the structuring element.
thresholdSeparates objects from the background by assigning all pixels with intensities within a specified range to the object and the rest of the pixels to the background. In the resulting binary image, objects are represented with a pixel intensity of 255 and the background is set to 0.
threshold intervalTwo parameters, the lower threshold gray-level value and the upper threshold gray-level value.
TIFFTagged Image File Format. Image format commonly used for encoding 8-bit and color images (extension .tif).
trainingIn OCR, the process of teaching the software the characters and/or patterns you want to detect during the reading procedure. See also reading and verifying.
truth tableA table associated with a logic operator that describes the rules used for that operation.

V

verifyingIn OCR, the process of comparing an input character to a reference character and returning a score that indicates how closely the input character matches the reference character.
vertical element spacingSpace, in pixels, between vertically adjacent elements. (OCR)

W

watershed transformA method of image segmentation that partitions an image based on the topographic surface of the image. The image is separated into non-overlapping segments with each segment containing a unique particle.
white reference levelThe level that defines what is white for a particular video system. See also black reference level.

WAS THIS ARTICLE HELPFUL?

Not Helpful