Digital Image To Extract The Ellipse With Linear Geometry

In the field of image recognition, the shapes, including lines, circles and ellipses etc., are senior medium that are generally used. Extracting the shape from given image is the essential step that every recognition algorithm takes. Lines, circles and ellipses are regard as the simplest geometric shapes, from which many more complex objects derived. Before recognizing the ultimate target, it avails to get simple shapes such as lines, circles and ellipses. This recognition process conforms with that of human beings who know things from simple ones to the complex. So this research of extracting simple shapes from images has a close relationship to practical applications.In present documents, the most useful method on extracting shape is Hough Transform. Analyzing a great deal of articles on HT and its improvements helps to build unambiguous frame about theories and methods of extracting geometric shapes.The task of this paper is that the objects in images, including lines, circles and ellipses, could be extracted at one time. The methodology used here distinguishes itself from HT because it can give several kinds of objects recognized at one time while HT only gives one.This paper's method gives a new approach that groups the pixels by larger-scale neighborhood and gets the parameters of the neighborhood by K-L transform. These parameters marked as x,y,A,0 , specifying the center coordinate, length and direction of the neighborhood separately, are the basic data for following process. Then some traditional algorithms of edge track and object classification are used to accomplish the task of gaining last objects.With interpreting the paper's method in detail, paper gives the deductive formulae and programs and trial results. These resources demonstrate the feasibility of this method. At the end, necessary analysis on merits and demerits is given.