Research On Fragment Reassembly Technology Based On Multi-feature Information Analysis

Fragment reassembly technology is one of the key technologies in the field of image processing.When the number of fragments,such as fragments of cultural relics,fragments of public security material evidence,fragments of corporate documents,etc.,is large,it will take a lot of time to complete the recovery task manually and it is easy to cause damage to the fragments themselves.Therefore,it has important research value to use computer technology for automatic fragment reassemblyResearchers both at home and abroad have done some research on the reassembly of two-dimensional fragments.Usually,fragments are divided into regular and irregular fragments and the reassembly task is completed according to different characteristics.At present,the results obtained by fragment reassembly algorithm are not often very ideal in many cases.For example,when using color features to restore some images with less color differentiation,or using texture features to restore some images with less texture differentiation,there will be more mismatched fragments in the reassembly process,which will increase the operation hours and reduce the matching efficiency.The commonly used image fragment reassembly algorithm based on contour feature is suitable for experiments with different shapes of fragments.For fragments with similar boundaries,it of ften causes a lot of mismatches,which affects the operation hours and the effectiveness of the algorithmIn recent years,in the research of image processing.the multi-feature analysis of image has made some progress.Therefore,this thesis makes full use of the multiple features of image to achieve fragment reassembly.The texture and color feature of the image are fused to obtain candidate matching fragments,which can effectively reduce the number of mismatched fragments and achieve the final global reassembly by using contour features.In addition,in terms of texture feature extraction,considering that the GLCM(Grey-Level Co-occurrence Matrix)texture feature extraction algorithm is suitable for a single type of image,a more applicable NGLCM algorithm is proposed.Experiments show that the algorithm proposed in this thesis can effectively achieve the task of fragment reassembly,and the operating speed has been significantly improved compared with other traditional algorithmsThe main innovation of this thesis is as follows1.Based on the GLCM algorithm,we proposed an NGLCM texture feature extraction algorithm suitable for color image.Six pixel matrices are built between different color channels of fragment image and three component maps of fragment are extracted.Then eight-dimensional texture feature vectors of six matrices and the eight-dimensional texture feature vectors of three maps are calculated to obtain the mean value of nine corresponding features as the final texture feature of each fragment.The Euclidean distance of the eight-dimensional texture feature vector between two pairs of fragments is calculated,and is used as the texture feature to obtain the candidate matching fragments.Experimental results show that the error number of candidate matching fragments by this method is significantly reduced compared with the traditional GLCM algorithm2.A new reassembly algorithm that fuses image color.texture and shape(contour)feature is proposed.The color and texture feature of the image are weighted to obtain the candidate matching fragments.The correct matching pairs are retained and the matching angle is obtained by contour matching.Finally,the matching-angle-based reassembly algorithm is used to realize the fragment reassembly.Compared with other traditional fragment reassembly methods,the proposed method can effectively reduce the number of cycles in global reassembly and improve the reassembly speed.