Research On Silhouettes-based Automatic 3D Reconstruction Methods

The 3D reconstruction based on single image is an important research content of virtual reality?computer graphics.It is a reverse engineering to recover 3D information from a single 2D image,having the ill posed problem of Polysemy.Three dimensional reconstruction can be done by contour,brightness,texture,light and shade,interaction,motion and zooming.But,low efficiency and low precision are still the key problems in the research.In this paper,it reconstructs the object from the single image which contains the side view of this object.An automatic 3D reconstruction method based on contour is proposed.Firstly,the input image is pre-processed,and then the corresponding component modeling algorithm is selected according to the different morphological features.The main work includes:(1)The preprocessing of 3D reconstruction based on contour is described in detail.By comparing the advantages and disadvantages of Graph Cut algorithm and GrabCut algorithm,we choose the better GrabCut algorithm for image segmentation;we proposed the resampling method of the contour points,based on Douglas Peucker distance interpolation algorithm;A fast thinning algorithm is used to extract the skeleton of the contour of the object.the preprocessing provides effective contour and skeleton information for 3D modeling.(2)According to the principle of human visual perception,this paper proposed a component-based modeling algorithm.We puted forward the idea of the control point grid modeling,by observing the morphological characteristics of the main components and combining with the human visual perception.Firstly,considering the contour curvature,the line length and the angle of nearest and intersection points,it choose the optimal point,and on the optimal line we set control point.After that we established 2D control grid.Secondly,based on the human visual perception,we constructed the three-dimensional lifting function for lifting the 2D control points grid,then optimizated and 3D triangulated the lifted control points grid;Because of there were great differences in the structure and morphology of the lateral limb components,It is divided into two types: symmetric streamline and general structure.Finally,the main body and the lateral component model were merged and the texture mapping was used to finish the modeling.Experiment shows that the component modeling method based on contour is efficient and has good modeling effect.(3)We build 3DCreator system platform to verify the above algorithms and techniques.The modeling results of several kinds of objects and performance test are given.The system integrated the CGAL platform,which can optimize and edit the 3D model.This paper proposed An automatic 3D reconstruction method based on contour,which can provide users with convenient,fast,WYSIWYG(What you see is what you get)modeling experience,and support for 3D printing,virtual reality,computer animation and other specific applications.