Fragment Splicing Method Based on Surface Texture Characteristic

Image Stitching Method of Bone Stick Fragment Based on Similarity Freeman Code Matching

“…If they are equal, the point is not a boundary vertex because its front and rear directions are consistent, and execute 1); If not, it means that the direction of the front and rear sides of this point changes, and execute 3); 3) Calculate the direction coefficients Lf and Rf on both sides according to equation( 5) and ( 6), and judge the relationship between them and the change threshold T according to equation (7); if both are less than or equal to T, and execute 4); otherwise execute 1); 4) Calculate the change error DVF of the current point according to equation (8). According to the vertex judgment rules above, if it is a boundary vertex, store the index value of the point in the set vertex and execute 5); if not, execute 5); 5) Add 1 to the index value of the current point to cover the current point with next pixel point，and execute 1) until all the points have been judged.…”

“…There have been many researches on image matching at domestic and foreign [1][2][3][4][5][6]. Wang Piao et al [7] proposed to achieve fragment matching by calculating curvature and least square fitting, extracting the surface texture features of fragments, and constructing constraints; GENG Guo-hua et al [8] proposed to estimate the similarity of feature points at the fracture of fragments, and used singular value decomposition to achieve accurate matching and stitching of fragments; Ji YUAN Jie et al [9] proposed to construct the bidirectional distance field of the characteristic points of the fracture surface, calculate the rigid body transformation matrix, and use the iterative nearest point method to achieve the stitching of fragments; Fang Ran-ran et al [10] determined the matching information of the image polygon vertices by calculating the length between the adjacent vertices of the polygon and using the improved "tank algorithm" to obtain the matching metric of the corner sequence. This method makes full use of the corner edge information of the outline but ignores the geometric characteristics of the curve, and the calculation amount is large when the contour representation and the vertex are matched; Xu Gang et al [11] proposed to select the feature points on the contour curve by calculating the curvature, and use the ratio of the arc length to the chord length of the feature segment to replace the characteristics of the curve for matching.…”

Research Progress in the Splicing and Restoration of Artifact Fragments Based on Point Cloud

“…For artifact fragments with rich and widely distributed texture information, Wang, Y., etc (Wang, Y., 2018) a method of first enhancing and then extracting the sharp features of bronze shallow reliefs using feature point projection distance. Wang, P., etc (Wang, P., 2018) calculates the curvature of surface points on terracotta fragments to extract valley ridge points and obtain the surface texture of valley ridge lines. Subsequently, analysis of regular textures to obtain cell textures and broken textures, and matching rules were developed.…”

“…When using surface texture for stitching, Wang P etc (Wang, P., 2018; based on regular texture to achieve splicing, and only regular texture on the armor parts of the terracotta warriors, for the rest of the parts, cannot be applied, Yang J etc is the combination of texture information and fracture surface information. The use of surface textures can achieve artifact fragment stitching, but has greater requirements for artifact textures and are not universally applicable.…”

“…Zhang H.C., etc (Zhang H.C., 2011) classified fragments into eight categories: concave, convex, ridge, flat, valley, narrow saddle, saddle, and flat saddle based on Gaussian Curvature and Mean Curvature. Wang, P. etc (Wang, P., 2018) analyzed the extracted texture information and classified textures based on intact and broken edges. Lu Z.J., etc (Lu Z.J., 2020) established a local reference system for the 3D point cloud for rotational projection to obtain a 2D image with depth information, and the rotational projection features were extracted, the Integral Invariant as locally significant features to complete the classification of different parts of the terracotta warriors using a combination of multiple features.…”