Geometry Image Generating Method Based On Subdivision

With the development of computer and network technology, the informationresources are growing rapidly.3D geometric model becomes a new generation of datarepresentation following the multimedia data model of sound, images, and video. Thedata of geometric model with high precision have a high requirement on networkbandwidth when transmission on the network. The method of geometry imagesignores the point connection of the original data, transforming the irregularthree-dimensional geometry data into regular two-dimensional image data. It meansthat the geometric data can be processed with the aid of the image encoding method,which is more mature. When storing and processing the geometry image transformedfrom the three-dimensional model, it is very important to make geometry imageoptimal approximation to the original data, especially for the model with lowresolution. In the process of parameterization, part of the three-dimensional model isstretched greatly, then some features are not well maintained when regular sampling.This paper researches on the geometry images based on the related method athome and abroad, intends to combines the subdivision algorithm on thethree-dimensional model and some error measure methods, adds the adjustmentprocedure to process the initial mesh, in order to improve the quality of reconstructionmodel. The method proposed in this paper is divided into two aspects:(1) Resampling strategy based on subdivision. The regular resampling can’t dealwith the area with tensile deformation and fail to capture the feature information,which will results in reconstruction error. Through subdividing the area locally andadding more points, the feature points on the parametric mesh will be also increased.The operation is equivalent to the local non-uniform sampling, which will improve thesampling accuracy on the region locally and obtain a new resampling result. Therefore,this paper introduces the interpolation subdivision algorithm to subdivide the originalmesh locally and obtain an adaptive resampling. Experimental results show thatcompared with the traditional geometry image, the proposed method can providebetter results on both PSNR values and reconstruction quality.(2) Adaptive geometry image generating method. Geometry image offers a simpleand compact way of encoding the geometry of a surface and its implicit connectivityin an image-like data structure. Unfortunately, part of the input model is stretched greatly in the parameterization process of geometry images generation. And theregular resampling fails to capture the features on the surface which will result inreconstruction error. This paper applies ICP algorithm, mean curvature and areastretching metric to determine the region to be processed. After dealing with the meshlocally, we can get a more accurate resampling result, which will produce higherquality reconstructions.