Video Object Detection based on Region Contrast

Abstract

Foreground object detection is one of the major impotent tasks in the field of computer vision which attempt to discover important objects in still image or image sequences or locate related targets from the scene. Foreground objects detection is very important for several approaches like object recognition, surveillance, image annotation, and image retrieval, etc. In this work, a new method was proposed for detection and segmentation foreground object from image or video in case when the targets are moving or not moving. The proposed method are able to detect important target for case the target is moving or not and can separate foreground object with high details. Comparisons with general foreground detectors such as background subtraction techniques were done and conclude that the proposed method has a high accuracy with detection rate 98.05.