Research On The Technology Of 2D Fragments Re-asembly Based On Contour Features

Fragments reassembling based on the contour feature is a central important topic in the field of computer vision,image analysis and pattern recognition.It also can be applied in many other fields,such as archaeology,art restoration,and molecular biology,and so on.The topic of fragments reassembling belongs to the fields of shape matching or curves matching.In this paper I have researched every stages of the two dimentional framents reassembling, binarization,contour detection,smoothing filter,contour points sampling,feature points detection,fenture representation,matching,reassembling.Mainly works as following.1.Presented a edge detection algorithm based on the judgment of the pixel points neighborhood.According to the characters of the binary images,the contour points' 8 neiborhood must have at least one point witch color is same as the background of the image.So we can judge the 8 neighborhood of the point in the image.This method can detect the contour of the fragment exactly,and its time complexity is as same as the gradient operator method.But actually we just need to judge the points on the fragment which in the image,and we don't need to judge the background points.It also don't need any math computing.So it can improve the efficiency of the edge detection.2.Presented a feature points detection algorithm based on the ratio between the arc length and the chord length of the curve local.In this paper,the feature points is the comer points which curvature is bigger.The common method is according to the curvature or the angle of a point's two direction's tangent,it requires a large amount of calculation.In this paper,we use the ration between the arc length and the chord length of the curve local to measure the degree of the blending.Taking a number of points along the contour's two directions from a point on the curve.According to the rules we cumpute this curve segment's arc length and the chord length between the curve segment's start point and end poin.Then we compute their ratio.We regard the local extremum points as the feature points.This can avoid when we use a single threshold to detect the feature points resulting in none of the feature points on the flat segment of the curve.This method doesn't require too much computing and easy to understand.3.Improved a multi fragments' reassembling method.Firstly match every two fragments,when we reassembled two fragments successfully,we regard them as one fragment,and then we modify the fragments' matching information which have matching relation with the two fragments.Rassemble them continuously until all of the fragments reassembled to one fragment or there are no correct matching.This method can reduce the times of the reassembling fragments' pretreatment,fenture representation and matching.Improve the efficiency of the multi fragments' reassembling greatly.The experiments show that these methods can ensure the correctness of the matching,and they also can reduce the computing of some steps.And only representing the feature points can prove the efficiency of the matching.All of these can prove the efficiency of the two fragments' reassembling,and alse can prove the efficiency of the multi fragments'reassembling.The multi fragments' reassembling method I proved can greatly prove the efficiency.