Tracking Objects in Video Frames with Moving Camera

Abstract

Image segmentation and tracking objects are essential in image processing used in segregating an image into regions of interest. Different kinds of methods have been proposed for image segmentation such as region growing approaches, watershed, and mean shift procedure. One of the challenges in image analysis, especially in tracking, is working on images taken by moving camera. The background in a moving camera constantly changes, and detecting moving object makes tracking very difficult. In this paper, the concept of edge detection, the histogram of simple and complex images, and similarity between sequences of frames are exploited to facilitate the process of segmentation. Comparing our hybrid algorithm with other methods reveals that our method shows less detection errors than the others with lower temporal expenses in performance