auto thresholding

[KieranJohn]

Hi,

I am using labviews auto threshold function with the clustering technique. The code is working nicley and I am getting good results. However I need to write a research paper and can't find the specific algoritm used. Is it the Otsu method or K-means method or something else entirely? I would be very greatful if someone could answer this question.

Many Thanks,

Kieran

[ChristopheC]

The algorithms are described in the Vision Concepts Manual that you can find here:

C:\Program Files (x86)\National Instruments\Vision\Documentation\NIVisionConcepts.chm

Here is what we have for the description of the Clustering algorithm:

Clustering

Clustering is the most frequently used automatic thresholding method. Use the clustering method when you need to threshold the image into more than two classes.

Clustering sorts the histogram of the image 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 repeats until it obtains a value that represents the center of mass for each phase or class.

Hope this helps,

-Christophe

[KieranJohn]

Thanks for your reply Christophe.

Its the Mathematical Formula that labVIEW uses to perform the clustering that I am looking for. I can't seem to find it anywhere.

Regards,

Kieran

[invancea]

Please have a look at

http://zone.ni.com/reference/en-XX/help/372916L-01/nivisionconcepts/thresholding/

The "In Depth Discussion" explains how threshold is selected based on histogram and provides formula for each thresholding algorithm.

[t_A]

I would like to know the way to implement the best method of adaptive thresholding for face detection in surveillance system. Which of the following auto thresholding techniques would suit my proj.

NI Vision has five automatic thresholding techniques.

• Clustering
• Entropy
• InterVariance
• Metric
• Moments

Apart from these techniques is there any other method that would suit my proj.

[ChristopheC]

It mainly depends on the content and histogram of the image, and what you're trying to extract.

- The clustering method is most often used when you need to threshold the image into more than two classes.

- Entropy is preferred when you need to extract a class with a small population, such as defects.

- Interclass variance is based on discriminant analysis. An optimal threshold is determined by maximizing the between-class variation with respect to the threshold

- Moments is suited for images that have poor contrast.

In addition, VDM offers local thresholding methods that you should consider if there is a light drift in your image (i.e. the algorithm does not apply a global threshold to the whole image, but instead works locally). This works well to isolate objects in images that exhibit nonuniform lighting changes.

I invite you to check the NI Vision Concepts manual for more detail about each algorithm.

You can use Vision Assistant to experiment with these different techniques and see what works best for your application.

Best regards,

Christophe

[Carmo]

Hello,

I'm, trying to use clustering method, but it's always returning only a binarized picture. I was expecting a picture with different levels like is explained here http://zone.ni.com/reference/en-XX/help/372916L-01/nivisionconcepts/thresholding/   In the clustering explanation with 3 classes ... I'm I wrong?

Thanks

[tiho]

Hi

I would like to ask if the tresholoding method used in labview is Otsu's adaptive tresholding because I saw somewhere that the Otsu's adaptive tresholding is equal to the clustering tresholding method.

Best regards