Tham khảo tài liệu 'active visual inference of surface shape - roberto cipolla part 5', kỹ thuật - công nghệ, cơ khí - chế tạo máy phục vụ nhu cầu học tập, nghiên cứu và làm việc hiệu quả | . Tracking image contours with B spline snakes 51 and with the snake accurately localising and tracking it. Amini et al. 3 have discussed the problems with the approach. These include instability and a tendency for points to bunch up on strong portions of an edge. They present an implementation based on dynamic programming instead of variational methods which allows the inclusion of hard constraints which may not be violated as well as the original smoothness constraints which do not have to be satisfied exactly . Their approach uses points on a discrete grid and is numerically stable. Hard constraints also allow the distance between points on the snake to be fixed and hence can avoid bunching. However the main drawback is that the method is slow since it depends on the number of the sample points and the cube of the search space. The B-spline snake A more economical realisation can be obtained by using far fewer state variables 184 . In 50 cubic B-splines 76 were proposed. These are deformable curves represented by four or more state variables control points . The curves may be open or closed as required. The flexibility of the curve increases as more control points are added each additional control point allows either one more inflection in the curve or when multiple knots are used 18 reduced continuity at one point. B-spline snakes are ideally suited for representing detecting and tracking image curves. Their main advantages include Local control modifying the position of a data-point or control point causes only a small part of the curve to change Continuity control B-splines are defined with continuity properties at each point Compact representation the number of variables to be estimated is reduced to the number of control-points 154 The B-spline is a curve in the image plane figure x s i where fi are the spline basis functions and Q are the coefficients or control points. These are positioned so that the curve locates the desired image .
