Machine Vision

Showing results for 
Search instead for 
Did you mean: 

NI Vision Concept Help Manual Proof Reading Or Errors

Go to solution

NI Vision Concepts Manual-June 2015, 372916T-01:


Hi everyone, here in this group knows about this manual, if not please right away have a look at this. It’s considered as one of the great manual for those who work on machine vision using NI LabVIEW, Vision Assistant and Vision Builder which all includes similar algorithms explained in this manual.


As I read through this manual noticed few typos, errors and mistakes. So, I just thought I would include my findings so that it can be helpful for next revision of the manual (If the suggestions/Corrections are valid).


Vision Basics Digital Images Image Types

-64bit Complex Color Color-32 bit floating for real part, 32bit floating for imaginary part 

Color Images

 -This representation is known as 4 ∞ 8-bit or 32-bit encoding (Shouldn’t be 4 x 8-bit encoding? Confused mainly about infinity ∞ symbol).

-This representation is known as 4 ∞ 16-bit or 64-bit encoding (Shouldn’t be 4 x 16-bit encoding?).

  Display Palettes

-In concepts paragraph: pixels can take 28, or 256, (should be 2^8 not 28).

-In Binary Palette, the periodic palette follows the values shown in table till g=6. After that it’s not following the according to the table values. 

Image Processing and Analysis

Image Analysis

Structural similarity Index

-structural information between the two images to computed(Compute) the SSIM value.

Feature Extraction:

-End of Gradients Computation some text hey is there, i couldn't find anything related to it.

Image Processing

Lookup Tables

-T(x) equation properly spaced, initially gives confusion because second and third line are not after the equal sign.

-In Predefined lookup tables, in table no commas given between log, power and next row also.

-In logarithmic and inverse gamma correction example: Power 1/Y example is wrong and copied from Power Y for Y=1.5 case.

Spatial Filtering

-In kernel definition- model b and d are interchanged and right topmost c should be –c instead of c.

-In Filter axis and direction-prewitt #2 kernel is wrong and it should be 0, 1, 1, -1, 0, 1, -1, -1, 0.

-In Edge extraction and edge highlighting- Prewitt 15 decomposition equation kernels are interchanged.

-This equation indicates that Prewitt #15 adds the edges extracted by the Kernel C (Prewitt #14) to the source image

-In Contour extraction and highlighting – modulus missing in equation when x>condition.

-This equation indicates that the Laplacian #2(#4) kernel adds the contours extracted by the Laplacian #1(#3) kernel to the source image.

-In Non linear Sobel filter equation- P(i,j+1)-P(i,j-1) should be multiplied by 2.

-In differential filter equation-third element is wrong and it should be P (i, j-1)- P (i, j).

Grayscale Morphology

-Dilation example image for both structuring element looks same.

-Proper closing equation- max(I,COC(I)) not OCO(I).


-In Concepts, Truth tables, Top row is b b it should be b= 0 1.

Frequency Domain Analysis

-In-depth discussion:


f(x, y is the light intensity of the point (x, y, and (u + v) are the


f(x, y) is the light intensity of the point (x, y) and (u , v) are the

-Inverse Fourier – f(x, y) instead of F(x, y)

-In FFT Display:

-Where R(u-> R(u,v) and I(u-> I(u,v)

Texture Defect Detection

-In What to expect:

-In Rotation Variation:If the texture under inspection can shift(Rotate) more than 5 degrees

-In Scale Variation:Texture defect detection is invariant to rotation of approximately ±10 degrees.

-similar to rotation and it should be invariant to scale variation of approximately 10%.

-If the texture under inspection can vary more than 10 degrees (%) in scale

Flat Field Correction In-Depth Discussion

-Estimation—Estimate the flat filed(Field) image using the Estimate Flat Field Model algorithm.

-Also Don't know why for this topic only Main topic name is also added unlike other chapters.

Particle Analysis

Image Segmentation


-Entropy equation: no negative sign and Pb(i) should be P(i) in first equation.

-Pb(w) is the should be Pw(i).

-In moments- kth moment mk of an image is calculated as

-In Global color Thresholding: Green Plane histogram is repeated as red, Green form 100-150 not 130-200 in description.

-In Local Thresholding: Thresholding

Local thresholding, also known as locally adaptive thresholding, is like global grayscale thresholding in that both create a binary image by segmenting a grayscale image into a particle region and a background region.

-Niblack Algorithm :In-depth Discussion: Setting k to 0 to(not required) increases.

Binary Morphology

-In structuring element: In square frame all are sqrt(2)d instead of only diagonals. Vertical and Horizontal neighbours should be d only.


In Primary Morphology Operations

-In table-2 , second column header should be After Dilation

-In thinning function-images A and B are interchanged.

-In thickening example-accoring to structuring element in c, (5 ,8) (6, 7) shouldb’t be gray as well? Means equal to 1.


Advanced Morphology Operations

-In border function: and their elimination them(not required) helps to

-In skeleton structuring element: what are ??

Particle Measurements:

-Center of mass x and y, ∑x and not subscript.

-Ellipses-a= ½ E2a , b= ½ E2b

-Ellipse ratio- no plus sign it should be equal sign.

-Ratio of equivalent ellipse axes = E2a/E2b

-Particle Measurements

Norm. Moment of Inertia xx

Nxx= no / (Ixx/A^2)

Machine Vision

Edge Detection

-When to use-In Alignment :Figure 11-3 no numbering?

-In Edge detection methods-Simple edge detection- A falling edge is detected when the pixel value falls below the specified threshold value minus a hysteresis value.

-Straight edge score: no equal sign in equation. s = c/m+n 

Pattern Matching

-When to use - Figure 12-1a and 12-1b no numbers?

-In Pattern Matching Tecnhiques-Gradient method- less data must be computer(Computed).

-In Tips and Tricks- fifth point -For larger templates with a well-define(defined) Region of Interest.

-In-depth discussion-equation-denominator-second one no power(1/2).

Object Tracking

-In object tracking techniques:Understanding meanshift: The mean shift algorithm is a is a(repeated) simple method

-In Background Subtraction- A second method used to improve the convergence of the target model the that(to) target object is called background subtraction

Geometric Matching

-In Geometric Matching Technique:In Generalized hough matching- point 2b- yc = y-r sin(theta).


-In many places, i saw You also can do, i feel you can also do would be appropriate.


-Apologies for bad formatting. 

-i am posting these corrections according to my knowledge if not correct feel freeto comment.



Message 1 of 5
indeed NI vision manual is not complete many of terminals of vision vi was lift without any definition of help some of them of them are vague
I think NI should work on this toolkit more I can help to improve it But some body from NI should start it
0 Kudos
Message 2 of 5

Agreed that the manual is incomplete and needs lot of work. Whatever is present at least should be correct without errors is what my intention of posting my findings.
-Mean time, if you want to suggest ideas you can post in here:

0 Kudos
Message 3 of 5
Accepted by topic author udka

Hi uday,


Many thanks for posting your findings! I will take this feedback to the team and see that these errors are reviewed for future revisions. 




0 Kudos
Message 4 of 5
Hi Joseph,

Thanks for the heads-up. Hope next version of manual will be error free.
0 Kudos
Message 5 of 5