how to select a seed point in clustering?

seed point selection in clustering technique for segmentation.

[Walter Roberson]

Are you asking how you tell a particular clustering routine which seed point to use, or are you asking for advice as to which location you should tell a clustering routine to use?

[Sreepriya Jeejesh]

i need a code for adaptive clustering.steps is here.could you please help me?

1. Define seed point Co by calculating the averageintensity of that image.
2. Define a pixels cluster which the intensities areless than Co
3. Calculate the average intensity C1 of that cluster.
4. Iterate the process by returning to step 2 fordefining additional pixels cluster and thencalculating C2.
5. We repeat above processes until(Ci-1-Ci)<T.where T is a calibrated parameter.

Co,C1,...Ck represent the cluster centers.

The final step is to group image pixels in such a way that pixel is assigned to the nearest cluster center measuring by Euclidian distance of intensity.

Answer 1

[Aditya Adhikary]

For k-means, you can specify the seed using the 'Start' parameter. If you specify a numeric matrix while using this parameter, it can interpret it as the seeds. For more information on how to use this option, read the documentation: kmeans Name-Value pair arguments.

[Sreepriya Jeejesh]

i need a code for adaptive clustering.steps is here.could you please help me? 1. Define seed point Co by calculating the average intensity of that image. 2. Define a pixels cluster which the intensities are less than Co 3. Calculate the average intensity C1 of that cluster. 4. Iterate the process by returning to step 2 for defining additional pixels cluster and then calculatingC2. 5. We repeat above processes until(Ci-1-Ci)<T. where T is a calibrated parameter. Co,C1,...Ck represent the cluster centers. The final step is to group image pixels in such a way that pixel is assigned to the nearest cluster center measuring by Euclidian distance of intensity.

[Aditya Adhikary]

You may like to refer to Adaptive k-means clustering for images.