Segmentation of Terahertz imaging using k-means clustering based on ranked set sampling

Terahertz imaging is a novel imaging modality that has been used with great potential in many applications. Due to its specific properties, the segmentation of this type of images makes possible the discrimination of diverse regions within a sample. Among many segmentation methods, k-means clustering is considered as one of the most popular techniques. However, it is known that k-means is especially sensitive to initial starting centers. In this paper, we propose an original version of k-means for the segmentation of Terahertz images, called ranked-k-means, which is essentially less sensitive to the initialization of the centers. We present the ranked set sampling design and explain how to reformulate the k-means technique under the ranked sample to estimate the expected centers as well as the clustering of the observed data. Our clustering approach is tested on various real Terahertz images. Experimental results show that k-means clustering based on ranked set sampling is more efficient than other clustering techniques such as the k-means based on the fundamental sampling design simple random sampling technique, the standard k-means and the k-means based on the Bradley refinement of initial centers.