Research On Fragment Reassembly Based On Edge Information

The broken file recovery technology is widely used in forensic science,antiquities preservation,medical and remote sensing science and other fields.For fragments with regular edges and less number,it is feasible to complete the reassembly by manual matching.But for irregular fragments,such as unearthed artifacts and broken pieces used for criminal investigation,which are mixed each other and have a large amount,it is difficult to get the original restoration objects simply by manual comparison.Also it is time-consuming and can cause secondary damage.Although the computer aided reassembly has less accuracy than the manual reassembly,it has high speed and high efficiency,thereby reducing the work intensity.Many fragment reassembly problems can be attributed to or are equivalent to the two-dimensional fragment reassembly problem.Research on two-dimensional fragment reassembly problem not only has a wide application prospect,but also has a strong theoretical significance.The two-dimensional irregular fragment reassembly can generally be divided into three steps.The first step is image preprocessing,i.e.,image analysis and processing,which can obtain the elements needed in the process of reassembly.The second step is fragment matching process,which uses algorithm analysis method to determine the correct matching pairs by finding the relation between fragments to be matched.The third step is the reassembly of fragmented images.The fragment pairs to be matched in above steps can be effectively stitched by a feasible method to finally complete the fragment recovery work.At present the research of fragment reassembly can be divided into contour-based reassembly method and content-based reassembly method.This thesis studied the two-dimensional fragment reassembly method based on contour information and put forward a new method by combining the edge matching and the intensity matching.Experiment analysis verified the feasibility of this method.The content and innovations of this thesis are as follows:1.The fragmented image processing basis is introduced.The pretreatment process offragments includes image acquisition and extraction of feature points and feature values.Since the number of pixels and the amount of data representing the fragment contours are too large,corner extraction method is used to represent the fragmentation contours and a simple internal tunneling method is used to extract the edge contours in this thesis.By marking the inner and outer pixel points the edge position is determined.Experiments show that this method is fast and accurate.2.The reassembly algorithm of two fragments is studied.This thesis uses fragment contours to find the initial matching pairs and proposes a method based on the gray-scale cross-correlation of edge length and fragment contour matching edge to finally determine the true matching pair.It can effectively eliminate the pseudo-matching fragments whose edge lengths are approximately equal.Experimental simulations show that this local matching method has faster matching and higher precision.3.The reassembly algorithm for multiple fragments is discussed.In this thesis,experiments with multiple irregular fragments are conducted and an improved depth-first traversal search method is used for global fragment reassembly.Simulation results show that this method is less affected by noise,and can be combined with the aforementioned partial matching method to obtain good reassembly results.