Motive Target Extraction And Enhancement In Surveillance Video

The video surveillance system lives everywhere in modern society. In the videosurveillance system, moving targets are usually focused on more than the background of themoving targets. Researchers are high on researching moving objects detection, movingobjects extraction and enhancement. Certainly, more reliability, better real-time and betterprecision for moving objects detection and extraction are the aims people striving to achieve.With these targets enhanced, beyond doubt, the video surveillance system would makedifferences on all aspects of our life.Firstly, the paper discusses the moving objects detection and extraction that the methodsof moving objects mainly detection depend on the motion characteristics of background andmoving objects. As moving object detection and enhancement I research on are on the staticbackground, the paper introduces tree algorithms–algorithm based on the difference theconsecutive frames, algorithm based on three-frame-differencing and algorithm based on thedifference from the background used to moving object detection and extraction on the staticbackground. When moving objects extracted by Algorithm based on the difference theconsecutive frames, the areas extracted would be double-shadows. Then, the algorithmbased on three-frame-differencing overcomes the shortcoming of double-shadows. Movingobjects area can be largely extracted by the algorithm based on the difference from thebackground. Then paper analyzes the advantages and disadvantages of them and come up anew algorithm based on the difference from the background based on Gaussian model.The paper discusses the algorithms of the enhancement of moving objects area. HEwhich is used to the whole image is harmful to the details to be conserved while LAHE whichis used to local image is helpful to enhance the details. But a fixed window in LAHE make nodifferences to enhance the details and sketches. The paper comes up a new LAHE with anadaptive window. It would be better than LAHE when dealing with the relation between localdetails and local sketches. Experiments show that the new algorithm enhances the PSNRvalue compare to the LAHE. That is to say, the new algorithm can find more information thanLAHE. The speed of image processing using the new algorithm is quicker than LAHE.