Object-Oriented Foreground Image Extraction

This paper proposes a novel foreground image extraction approach, it elaborately combines two kinds of methods (foreground extraction and object detection) to obtain more accurate and more complete foreground object image. The proposed method consists of three main processing modules: frame-based distinct pixel extraction (FDPE), object detection (OD), and region-based foreground object extraction (RFOE). FDPE points out the distinct image pixels which have quite different features from their background models. When one region contains enough distinct pixels, OD then evaluates whether this region contains an instance of one specific interested object or not; and RFOE extracts the foreground pixels from regions containing the interested objects. The effectiveness of our method is demonstrated by means of experiments.