Research On Image Stabilization And Target Tracking Technology Based On Video Image Sequence

Nowadays, the video sequences stabilization and target tracking technology become one of the most popular topics. With the increasingly widespread used of mobile electronic devices, handheld cameras, cameras and other ship-board, video captured by camera movement exist jitter. Jittering video not only affects the visual effect of watching, but also cause difficulties for post-processing, such as:target tracking, video compression. Therefore, the stabilization of the irregular video jitter is an important task. Target tracking technology is widely applied in video surveillance, human-computer interaction and other areas, this paper also have some research on target tracking.There are two main units in digital image stabilization system:the motion estimation unit and the motion compensation unit. The motion estimation unit is more important, its purpose is to estimate the reliable global camera motion through the acquired image sequence via diverse algorithms. We study several usual motion estimation algorithms, then propose a novel one based on image features. Firstly, we extract the points which have obvious texture in the image. Secondly, form a feature block centered at the feature point, and search for optimal matching block, then using Distance Constraint reject the points that are incorrectly matching or on motion objects. At last by solution the motion model we can get the global motion parameters. Motion compensation is to get motion compensation vector by motion filter, according to the motion compensation vector, move to the opposite direction to compensate for the image to remove jitter components, we study and use proportional-integral controller for motion compensation. In target tracking, this paper in the framework of particle filter, propose an improved color space model as the observation model to calculate the Bhattacharyya distance between target template color histogram and particles color histogram to determine the exact location of the target. At the histogram statistics, according to the principle of the integral image, we use fast integral histogram method to statistics integral histogram the area of searching. To calculate the target candidate histogram, only need addition and subtract operation. Experimental results demonstrate that the propose stabilization and target tracking methods provide robust and good results, can be on-line.