Based On Images Of Fast Rendering Technology

Image-based rendering (IBR) techniques have been proposed as an efficient way togenerate novel views of real and synthetic objects. With traditional gcometry-based renderingtechniques, the time required in rendering an image increases with the geometric complexity ofthe scene. The cost of IBR techniques just depends on the resolution of the images and it canoffer an easier modeling process at a higher realistic level. Since its initiation in the early l990s,IBR has been developed very fast and widely used in many areas such as virtual reality computeranimation.This thesis restricts itself on the topic of image-based rending techniques. Some efficientIBR algorithms have been proposed to render novel views of complex virtual environmentinteractively, or in real time.In Chapter 2, we present an efficient depth-estimating-based IBR algorithm for novel viewgeneration from multiple reference images in a static scene by combining forward mappingtechnique and inverse mapping technique. The method successfully fills the holes in the derivedimage in both cases where object surfaces are magnified in the novel view or occluded surfacesbecome visible. This algoritlun is more efficient by the heuristic depth esthoating techniqueproposed in this paper which gets displacement values used during backward mapping processfrom a single primary reference image under the observation that, the shape of space surfacesoften changes smoothly In comparison with usual forward mapping methods, this algorithmgenerates derived images with less errors and tackles the problems associated with multiplereference images.By taking advantages of epipolar line features and depth discontinuities in reference 中国科学院软件研究所博士学位论文 基于图象的快速绘制技术的研究 images,an efficient inverse wmping algorithm is pfoposed in Chapter 3 for gCnerating nagCs of novel views by combining multiple eference images帆enhm different vie呷oints．Because continuous segnents determi。d by pairs ofedge pixels at co。spending epipolar lines are order－kept,only pairs of edge pixels in the reference渝明e e necess叨口cowute to obtain generalized disparity of all pointS in the desired image．As a result,sighficant acceleraion could be made in the endering pfo比鸵．Two accelerating techiq此s e presented in this algori山mb accelerate the hole illing process．his algorithm extends the reference images rom projection of single col:II'－ected surface in previously develoPed nvnverse w出下er to ima驴s caPtUred rom complex scene． In Chapter 4,an《dent IBR method is prese庇仙 y takn ull讪antage of呷bies c咖． The method can simulate the 3D details on sllri:ace of Object successfully．he。叩proach, called rered ature mopmp consists of two pans．At fst,an origi。ltextUre with orthogonal displacements per pixel is deCO啊osed into a series of new t6刀mfCs with each他lug a given displacement per pixel, called ae,ea atures,or LT hen． hese LT e used to render the novel view by conventional texture mapping．D avoid gaps n the endered hlla驴,some phels are to be interPolated nd extended in the厂kaccoding to the depth differe eee between two neighbor pixels In the original texture．As these LTLT fC…e much storage nd therefore much time is equired to install LTLT into the text'UfC buffec an旧thod is pfoposed to co呷fCss the LTLT, nd the cotTCspondingfClldering method is given．Experimental esults show that the new method is efficient,especially n rendering those objects with a smaller depth rnge compared withtheir size, such as relief surfaces of building．...