Efficient region boundary approximation using adaptive smoothing and second-order B-splines

Boundary encoding can be a key part of a content-based image retrieval (CBIR) system. An efficient algorithm is presented for approximation of image region boundaries by piecewise second-order B-spline curves. The algorithm first applies an adaptive low-pass filter to smooth the boundary while maintaining a specified tolerance of positional error. Next, critical points are identified and utilized to partition the boundary into segments. The algorithm then approximates each segment with a second-order B-spline that meets the error tolerance. This algorithm is much more efficient than previous methods. Experimental results are presented using various boundaries extracted from images