Study On Key Technology Of Image Based Rendering

Image Based Rendering(IBR) techniques have recently received much attention in the Virtual Reality(VR) and Graphics community because of their potential to create very realistic image on popular hardware platform. The dissertation addresses the following several key techniques: the image mosaic, the image morphing, the simulation based on image, hybrid render based on geometry and image.1. On the research of the automatic image mosaic, we present a new automated panoramic mosaic algorithm based on feature. The algorithm solve the problem how to decide the coefficients of the initial, the weighted function and the width of the transition zone. The algorithm can be finished with automating in most case and suits for any scale image with unknown any parameter of camera. Mosaic image is smooth and has no gap.2. A new image morphing algorithm based on Bezier feature curve is presented, which specifies morphing feature using Bezier feature curve. Compared with field-based image morphing, we can specify the feature more exactly and the numbers of feature primitive are reduced. From experimental results, the algorithm's speed is treble improved generating a smooth morphing image, comparing with field-based image morphing.3. A automatic image morphing algorithm based on wavelet edge detection is presented. The algorithm can automatically specify morphing feature. The method modifies field-based image morphing, reduces the complex degree of morphing algorithm and accelerates morphing. From the experiment results, the algorithm can automatically metamorphose image and avoid the most tedious part of image morphing.4. A automatic image morphing algorithm based on modifying active contour model is present, which automatically specifies the feature point with modifying active contour model. From the experiment results show that the algorithm reduces the burden of feature specification and metamorphoses automatically.5. We present a new algorithm of facial expression animation based on radial basis function networks. The algorithm uses radial basis function networks to construct the warp function using only a small number of anchor points. The animation result shows the algorithm provides a powerful mechanism with processing facial expressions and suits for general hardware platforms and Internet environment.6. We present a new hybrid render algorithm based on geometry and image. The algorithm divides the scene into foreground and background. Thebackground is rendered with panoramic image based on multiresolution spline and represented as multiresolution level, so as to improve the speed. The foreground is realized with a simplification algorithm generating different level of detail 3D facial model based on the edge collapse, so as to meet real-time rendering. The algorithm is widely used in many areas, comparing with Debevec algorithm.