Salient Feature Extraction Of3D Meshes And Its Applictions

With the ever-growing of the quantity and quality of3D mesh model, and its widespread using on human-centered visual computing applications, the theory of human vision perception will be introduced into3D mesh applications. The salient feature extraction of3D meshes includes salient regions extraction and feature lines extraction. Here, the robust definition of visual saliency measure is proposed and its application on viewpoint selection of3D meshes is given. A3d sculpturing technique by using user defined sketch lines is presented and an algorithm for the loop feature lines extraction is proposed. These issues become a significant research in the field of digital geometry processing, computer vision, etc.The main work and achievements in this article are as follows:1) To automatically select the best viewpoint of3D models, a relief saliency entropy driven view descriptor is presented, which is defined by incorporating the relief saliency measure of complex models into the entropy analysis of viewpoint mutual information. The saliency information entropy driven view descriptor is calculated for each pre-sampled viewpoint on the viewing sphere bounding the underlying3D object. Thus, the viewpoint of the highest value of view descriptor is the one that defines the best view.2) Due to the user defined sketch lines on3D triangular meshes, a novel3D sculpturing technique is presented. Firstly, we delete the triangles from original model which through by sketch lines and reinsert new triangles to original model after subdividing the deleted triangles. And then we calculate the distance between vertices on model and user’s sketch lines, determining the vertices that the distance is not longer than defined threshold. We also compute the shift distance and direction of vertices based on user’s line width and crease profile. Finally, we change the position of vertices near the sketch lines on the model for sculpture result.3) Based on the feature point set evaluation scheme, a fast algorithm is presented in this paper to extract the loop feature lines of triangular mesh. Firstly, the feature point set can be determined by the concave-convex property of their neighboring sampled points. The initial feature point of loop feature line can be selected automatically or given by the user, and the next feature point can then be iterated along the principal direction of the current feature point. The iteration step will be continued until the final feature point meets the initial feature point. Finally, the loop feature lines can be extracted by using the cubic B-spline curve interpolation scheme on the extracted feature points and by projecting the curve onto the underlying mesh.Both the visual saliency definition and feature line extraction are some basic researches related to the salient regions of3D meshes. Meanwhile, the best viewpoint selection and3D sculpturing technique are some applications related to the feature regions of3D meshes. Based on the proposed visual saliency measure, we will improve the best viewpoint selection scheme and the corresponding determination of best view direction in the future. Moreover, we could improve the3D sculpturing results by optimizing the vertex shift direction and by using two-dimensional sketch curve projection onto3D meshes as a new interactive manner.