How to estimate the the optimal number of clusters in a dataset by Clustering quality measures?

Question

How to estimate the the optimal number of clusters in a dataset by Clustering quality measures?

I have four datasets iris,breast cancer, magic ,wine and yeast. All the datasets are taken from UCI repository.

Am I supposed to run Kmeans or Any hierarchical algorithm with some k to estimate the number of clusters ?

Below is an example from a research paper. 6Gauss , Paired,Elong,Uniform and Rings are the datasets.

http://postimg.org/image/65x2ff8ph/

There the researcher estimated the number of cluster estimations in each Clustering quality measure.

Here is the link to the original paper