Localized Salient Region-Based Object Detection and Tracking

This paper proposes a method to embed multiple features of an image for object localization and salient region for object detection and tracking. Initially, the object location is obtained using ground truth and features are extracted. Subsequently, these features are used to find the potential area of an object and this area is used to identify a salient region. The pixels color and gradients are used as feature matrix and saliency maps are generated using the visual cue like the local contrast. The salient region detection has proved its application in object recognition and segmentation. The combination of feature matrix and saliency map enables the proposed tracker to be robust. The proposed method efficiently localizes the object even when there are scaling, rotation and illumination changes. The experiments are conducted using publicly available Visual object tracking (VOT) 2016 dataset which consists of many challenging video sequences and the proposed method provided competitive results when compared to many state-of-the-art methods which is evaluated using the Visual object tracking toolkit.