Corner detection and curve representation using cubic B-splines

Local Invariant Feature Detectors: A Survey

“…Several techniques have been developed which involve detecting and chaining edges so as to find corners in the chain by analyzing the chain code [205], finding maxima of curvature [108,136,152], change in direction [83], or change in appearance [42]. Others avoid chaining edges and instead look for maxima of curvature [254] or change in direction [104] at places where the gradient is large.…”

Reverse Engineering of 3D Models Based on Image Processing and 3D Scanning Techniques

“…From the discrete representation of the 2D boundary we derive a continuous B-spline curve representation according to [10]. The B-Spline approximating the boundary of the object is superimposed on the intensity image of the object to be modeled.…”

Identifying Salient Circular Arcs on Curves

“…[1,7,11,15, 23], they do not lend themselves to the description of curving segments of contour as units unto themselves, as seen in figure 2. For this reason, workers in geometric modeling as well as computer vision have turned to more complex parametric models, including circular arcs [6,10,16, 24J, more general conics [5, 271, and splines [12,14,19,25]. In general, the analytic form selected for describing contour segments should be matched to the domain-dependent processes that generate the contours; for example, if all images for a given task are oblique views of circular objects, then elliptical models are appropriate for describing the contours which will be found in the resulting images.…”

“…Most previous work with curve segmentation, including segmentation in terms of circular arc approximations, treats the problem as one of finding some optimal set of "knot" points which decompose the contour into disjoint segments that meet end to end [1,7,10,12,15,19,30]. This approach is well suited to the problem of reconstructing the original contour from an information-compressed representation.…”