Moving object tracking in the sequence of images acquired from non-stationary camera

This paper discusses the possibility of moving object tracking in the sequence of images acquired from a non-stationary (even shaky) camera that is held in the hand. In the process of image acquisition the effort is to keep the moving object in the central area of the image. In this way we manually prompt the evaluation system to recognize and consequently track the object concerned. This task is very complex because the camera is held in the hand and the object is moving far from the camera. In addition to the shaking camera, we do not have any a priori information about the scene and the objects detected. This leads to a very shaky sequence of images. Such a problem has not been studied systematically. The proposed method is based on the weighted gradient object edge detection and object size and color tracking. To speed up the evaluation process for real-time applications no previous image stabilization has been processed. The method is demonstrated on a sequence of images acquired from a real situation.