Algorithm of Clustering for Color Images Segmentation
P. Martinez Galaviz, A. Rojas Lopez, Marc Garcia, Ivo Torres · 2005
The problem of recognition of objects inside an image can be divided into two steps: the segmentation phase and classification of the objects. The cluster is a representative method of the segmentation technique that divides the pixels into different groups, based on properties of coherence and similarity. We propose a variation of the implementation of the cluster method based on density, OPTICS, which considers creation of clusters and the density of the points and develops a classification with regard to their parameters. In this method, we choose the color attribute as an important element to define a cluster; the color space considered is the HVC. The experiments were carried out with images that consider different situations; in most cases the results are close to what we expected, once a cluster is found the classification process can be done.