The Object Abstract Of Snake Model Based On The Seed Filling And Tracting

Image segmentation is one of important research field in computer vision and patern recognition.Beause it plays more and more important roles in many fields, many methods about it come out. Active contour model is a top-down processing procedure with high level information. Via this method, extract object or locating feature of interest in images becomes more efficient and correct.Subject to internal,image and other constraints, active contour models are deformable curves using energy minimization. The main limitations of the models are that they are sensitive to noise and pseudo edges; and that they must be initialized close to the feature of interest. Aiming at theses problems, active contour models based on region information are important orientation of technological development.In this paper, we will summarize the strengths and weaknesses of the original model on the base of discussion in detail, and synthetize ideals that proposed by many scholars, then ameliorate the model from two aspects: On the one hand, we discuss the structure of the energy function and introduce the Gradient Vector Flow to the energy function to improve the components of the energy function. We use Gradient Vector Flow and overcome the shortcoming of deep contour, then we use the sobel operator which is improved to detect eight direction, which is more robust to noise,and the snake model could exactly distinguish all kinds of contours. we can use the snake model to detect the image edge which has more serious noise and more complex texture. The image contour which is detected by the eight direction sobel operator and the image contour detected by the GVF snake model will be the object cuntours. On the other hand, seed filling is adopted to extract object contour in the paper and then segment object contour, and then the extracted object contour is taken as initial contour of snake model for precise segmentation computation. So it improves convergence speed.Finally, morphology processing was performed on the binary image obtained from the previous step to get the segementation result. Experiment results show that the new algorithm can segement video objects accuratelt and automatically. Hausdoff_based tracker and snake model are used to realize object extraction in subsequence frames.The Hausdorff measure that incorporates both location and orientation information is used in model matching and model updating process in the techniques. A high efficient searching strategy in location object in a image has also been proposed. Results indicate that these techniques can perform well even under dificult condition.With the research of segmentation and tracking of video object, we take note of algorithms applicability and relation between segmentation and tracking. In the strategies, author adds preprocession and post processing to ensure the validity of the algorithms.