The Method Of Three-Dimensional Model Transmission Based On User Specified Reconstruction Error

As users’ demands of vivid3D models which are used to applications such as online games and distance education gradually increase, the datum of all kinds of models increase sharply. In order to make the3D model transmit on the limited network bandwidth in real time better, the technology of3D model compression has become a hotspot of computer graph.High-pass quantization compression coding technology transforms the Cartesian coordinates of the mesh to relative δ-coordinates. Although quantizing the relative δ-coordinates which cause the data loss resulting in low frequency error of the reconstruction model, compared with the original model, the reconstruction model may slightly shift, scale or rotate which is caused by low frequency error, but the local features and curvature are well maintained. The error will not give rise to visual disturb. In this paper, we do some work which is based on High-pass quantization compression coding technology, the main point is as follow:1) We proposed a method of user specified reconstruction error for3D model transmission based on Laplacian coordinates. This method allows the user to specify the visual error level Sq and the low frequency error Mq of the reconstruction model. Firstly, the server-side transform the Cartesian coordinates of the mesh to relative8-coordinates, according to user’s visual error level Sq, we quantize the relative8-coordinates with corresponding precision. Secondly, we record the concrete anchors at different level of parameters v (parameter v denote dense degree of anchors). We match the curve about low frequency error and parameter v:Mq=f(v), on the base of its inverse function v=f-1(Mq) and low frequency error Mq specified by user, we can know the quantity of anchors. Finally, using Greedy algorithm to pick a few anchors to adjust actual low frequency error of reconstruction model.2) Base on the experimental result we draw a conclusion that the quantization accuracy of relative8-coordinates has a crucial influence on visual error of the reconstruction model, With the increase of quantitative accuracy, the visual error of reconstruction model is reduced to half as the quantitative accuracy per-bit increase. We decide the visual error of reconstruction model through controlling the quantitative accuracy of the relative8-coordinates.3) We have discussed the relationship between low frequency error of reconstruction model and dense degree of anchors. This method has a unified standard for3D models whose vertices numbers vary widely to measure the relationship between low frequency error of reconstruction model and anchors. From experiment we found that the function Mq=f(v) has two kinds of forms:linear and exponential type, but as the quantitative accuracy of relative8-coordinates raise, the form of Mq=f(v) will change from linear to exponential type, that is to say the higher quantitative accuracy of relative δ-coordinates, the lower effect on low frequency error of reconstruction model.