Research On The Image Segmentation And Shadow Removing Algorithm

In the field of intelligent video monitoring, screen and video technique multimedia application technique, the body or others must be detected and separated from the background. That is the problem of segmenting objects from real-time background. The segmentation result of vedio image is with important influence to the post-treatment of object sorting, tracking and behavior understanding.Image segmentation has been attached importance to by researchers and thousands of algorithm have been brought forward. Although these algorithms are with success to different degree, there is a long way to solve the problem of image segmentation and the research is still faced with a great deal of challenge. Nowadays, the popular methods of image segmentation often ignore the shadow detection. The shadow is detected with the object. Shadow causes severe problem such as object uniting, object form distortion and so on, which cause error to segmentation and tracking.Firstly, The theoretical basis of image processing is discussed. The conventional edge-detecting operator is theoretically analyzed and tested with simulation experiment. The experiment result indicates that while the transition of the edge gray value is aculeated and the noise is small, the effect of gradient operator is perfect.Secondly, several color space and the conversion algorithm between them are introduced. The area of application and disadvantage as well as advantage are analyzed. According to the shoot condition and aimed to the concrete application request, the image segmentation is accomplished based on RGB space. In the RGB color space, the image is firstly segmented based on the improved method of difference between the foreground and background image. Then, the noise is restrained through the process of region growing. The image segmentation result is that the shadow is segmented with the object.Thirdly, the basis hypothesis and general frame for the shadow detecting is summarized and analyzed. The main shadow detection and removing algorithm at home and abroad are discussed. Also, the disadvantage of each algorithm to remove shadow is pointed out.Finally, aimed at the characteristics both of the shadow and object in different image, four kinds of algorithm based on the basis hypothesis and general frame for the shadow detecting are designed. The shadow removing algorithm based on edge information is aimed to remove the shadow casted by such object as the one with rich edge information. The algorithm based on RGB or HSV color space are both applicable to the shadow that is with different color information to the object, and while shadow is with the same color information as object, both of the two algorithm are failing to remove shadow. To solve the problem above, the fourth algorithm based on HSV color space and the first gradient which is aimed to remove the shadow with similar color information as object, is brought forward. Experiment results indicate that the fourth algorithm is comparatively ideal among the four algorithms.