A Comparative Study of Clustering Validation Indices and Maximum Entropy for Sintonization of Automatic Segmentation Techniques

Automatic image segmentation is a fundamental task in many applications such as video surveillance, image retrieval, medical image analysis, recognition, tracking and objects classification. This task is not easy due to the complexity of image features as well as the number of objects within the images which is unknown in most of the time. However, the correct tuning of the parameters associated with the automatic segmentation algorithms can improve the levels of precision in the object segmentation in the images. This paper proposes the use of clustering validation indices and maximum entropy as cost functions to quantify the quality of segmentation in order to find the optimal parameters of the automatic segmentation techniques. The impact of using clustering validation indices and maximum entropy was evaluated in the automatic segmentation algorithms, K-Means, Watershed and Statistical Region Merging (SRM), using four image databases with images containing different numbers and sizes of objects and different levels of illumination. The results obtained reveal that the use of clustering validation indices and maximum entropy for tuning the automatic segmentation algorithms results in better segmentations with a number of segmented objects close to what really exists in the images, while these results are competitive with the works reported in the literature.