Relative object localization using logistic regression

This paper discusses the technique of relative localization of object (Region Of Interest) based on logistic regression classifier. In proposed method, local features for the whole image computed and represented as a set of vectors for training the logistic regression classifier to classify the given feature vector either to the region of interest or the background. With the classification, the background and the region of interest can be separated. Then, the region of interest is localized with due reference to the background in the subsequent images. The method shows advancement in the object localization and it additionally works in real time without being the computationally comprehensive approach.