| The fast splicing matching algorithm for irregular edge graphics are studied in this thesis.The traditional fat splicing and matching algorithm of edge graphics usually adopts Freeman code or morphological-based algorithm.Freeman can simplify the edge description on the edge representation of the image,and transform the edge features into digital characters to describe edges.But Freeman code has redundancy in the representation of image contour,insufficient processing ability in graph rotation and blurred boundary.Morphological algorithm can quickly realize the mosaic matching of edge graphics.However,morphological algorithm is very sensitive to jump points and noise points on edge graphics.In order to overcome the shortcomings of the above methods,this thesis proposes an edge description method combining edge coding and graphics morphology algorithm,a polar coordinate edge coding method based on disk model,which is a disk model established by itself.The description information of the graphic is obtained by using the position information of the model boundary.The feature acquisition of diagonal point,tangent point,inflection point,endpoint,breakpoint,line representation point and curve marking point is accomplished by using the description information of graphics.The detection and classification of two kinds of coding points are completed.Finally,the scale feature is added,and the graph restoring ability is increased.Seven or eight kinds of features are encoded for different calculate requirements.The coding method under the new model preserves the necessary feature points and encodes the codeable points,removes the unnecessary redundant feature points,reduces the code length,and identifies the graphic form by the combination of the feature points.Thus,the disadvantages of the above methods are overcome.The main work mainly includes the following four aspects:(1)Establishment of coding model based on polar coordinates under disk model.The maximum Euclidean distance of the contour of the target figure is obtained by obtaining the high quality edge contour of the original image.The disc model is established as the diameter of the disc model,and the coordinates of the boundary pixels of the figure are converted to the polar coordinate system.Using the characteristics of the disc model can quickly determine the line and curve marking points of graphic features;polar coordinates are less sensitive to deformation than other coordinate systems,which is conducive to recognizing spatial targets,and it is more convenient to calculate the line marking points and curve marking points.(2)Seven kinds of feature points construct two kinds of coding points and encode them.Seven types of feature classes include corner points,tangent points,inflection points,endpoints,breakpoints,line marking points and curve marking points,which can better describe the boundary of the graphics contour.Two types of coding points,i.e.feature pixel points and maximum coding length endpoints,complete the initial coding according to the established coding principle and obtain the original code.(3)Establishment and verification of fast stitching matching algorithm.In order to overcome the sensitivity of original coding to the selection of coding points and the variability of rotation,the normalized difference codes and complements are obtained by using normalized difference calculation and complement calculation.The normalized difference codes and complement space and their corresponding sub-codes space are established,and the mapping of sub-codes space is used to complete the fast stitching matching calculation.(4)Selection of evaluation criteria for algorithms.There are many evaluation criteria for mosaic matching algorithms.In addition to the speed of mosaic matching,the adaptability and error rate of the algorithm should also be considered.This paper chooses compression ratio,error rate and stitching matching rate to evaluate chain code algorithm comprehensively,and gives the deduction process in chain code algorithm. |
PDF Full Text Request