The Study On Fitting Image Contours With Bezier Curves

Curve fitting has many applications in image processing, computer graphics, pattern recognition and computer aided geometric design. Curve fitting is one of the key techniques in computer graphics, many domestic and foreign scholars have made a lot of research in curve fitting. Bezier curve is widely used in computer graphics to model a smooth curve, Since Bezier curve has excellent properties. The study on using the Bezier curves to fitting discrete points has important theoretical and practical significance.Curve fitting of image contours means representing the image contours into the form of a curve or curves. This paper will use Cubic Bezier Curves to fit the image contours, which needs finding the feature points of the image, and these points can reflect the image’s shape information and matche the image reliably. Because corner point is an important local image feature, has a lot of important shape information of the image, and can be reliably matched the image, so we choose corner point. The specific approach to fitting the image contours with Cubic Bezier Curves is firstly detecting the image’s corner points as the curve endpoints, then constructing the control points of the curves, and lastly using the endpoints points and control points to draw the contour curves. The choosing of the image corner detection algorithm and the method of constructing the proper control points are the keys to obtain high precision fitting contours.This paper firstly puts forward the concept of curve quantile, introduces the Adaptive Corner Detector Based on Curvature Scale Space, and modifies it slightly to make it have the ability to extract the curve quantiles of the contour curves between any adjacent corner points; and then introduces the Bezier Curve, studies the necessary and sufficient condition of G1 continuity between two Cubic Bezier Curves; finally presents a algorithm called fitting image contours with Cubic Bezier Curves:firstly using the modified corner detector to obtain the corner points of an image and extract the curve quantiles of the contour curves between any adjacent corner points, then use these points to construct the control points of segmented Cubic Bezier curves, finally draw these segmented Cubic Bezier Curves in MATLAB. The experiment results the algorithm can vectorize the image contours and has a continuity.