Robust Method for E-Maximization and Hierarchical Clustering of Image Classification

Abstract

We developed a new semi-supervised EM-like algorithm that is given the set of objects present in each<br />training image, but does not know which regions correspond to which objects. We have tested the<br />algorithm on a dataset of 860 hand-labeled color images using only color and texture features, and the<br />results show that our EM variant is able to break the symmetry in the initial solution. We compared two<br />different methods of combining different types of abstract regions, one that keeps them independent and<br />one that intersects them. The intersection method had a higher performance as shown by the ROC curves<br />in our paper. We extended the EM-variant algorithm to model each object as a Gaussian mixture, and the<br />EM-variant extension outperforms the original EM-variant on the image data set having generalized<br />labels. Intersecting abstract regions was the winner in our experiments on combining two different types<br />of abstract regions. However, one issue is the tiny regions generated after intersection. The problem gets<br />more serious if more types of abstract regions are applied. Another issue is the correctness of doing so. In<br />some situations, it may be not appropriate to intersect abstract regions. For example, a line structure<br />region corresponding to a building will be broken into pieces if intersected with a color region. In future<br />works, we attack these issues with two phase approach classification problem.