MIX RENDERINGS TO FOCUS PLAYER'S ATTENTION

Transcription

1 MIX RENDERINGS TO FOCUS PLAYER'S ATTENTION Vincent Boyer, Jordane Suarez Kamel Haddad L.I.A.S.D. Université Paris 8, 2 rue de la libérté Saint-Denis, France boyer@ai.univ-paris8.fr [dex, pepix]@bocal.cs.univ-paris8.fr KEYWORDS Rendering, GPU programming, NonPhorealistic Computer Graphics. ABSTRACT We propose in a same frame in a video game focus player's attention. Mixing kinds in same frame permits enhance informations on 2D images or 3D scenes. Images in video games are essential are main vecr informations for player. With our model, in high-scale scene, some informations can be enhanced. Our model allows designer choose used simultaneously; build GPU program that ses s export it. The generated GPU program achieves scene with selected by designer in one pass. This new GPU program can be imported by designer in video game can be used while key points are given. Or values, like associated distances each technique used can be changed dynamically in video games without any changes in GPU program. Examples are given images produced by our system are discussed. INTRODUCTION Images are main informations vecr in video games. Generally, one only one is used in a video game. For example celshading (Lake 2000) is used in GTA, phorealistic with High dynamic range imaging Tone mapping (Mantiuk 2004) are used in or car video games. One main ideas paper is consider that we can use simultaneously more than one technique at each frame render scene in video games. Many can be used simultaneously: artistic shading like Hatching (Winkenbach 1996), (Hertzmann 2000); deformation 3D mesh according viewpoint (Rademacher 1999); bump mapping, wood texturing (Rost 2004), classical gouraud shading, texturing... As one can imagine, this can be used focus player's attention on a path or an objective. Or it can be used create more complex level on a video games. In computer graphics this idea is recent but in architecture for example, this technique is commonly used create architectural models

2 in city planning: new buildings are designed with details (colors, windows, trees, flowers) while or existing constructions are designed with ir shapes only. We have proposed a method (Boyer 2007) for 3D scene visualization. The used are imported by designer a GPU program is aumatically dynamically generated selected. Based on GPU programing, this model is real time for large scale scenes visualization. In following, we present general algorithm our model with aumatic generation GPU programs. Then we present method used optimize export GPU programs. Examples on video games are n detailed. OUR MODEL In this section we describe briefly model used (Boyer 2006). The description is given introduce terms used in following. In this model, user specifies: 1. key points which are points 3D space. In following n is number key points given by user; 2. used s for each m four distances d0, d1, d2 d3 where [d0, d3] is range within is used : d0: minimal distance in order use this ; d1: minimal distance where we maximize use this. Between d0 d1 will be shaded; d2: maximal distance where we maximize use this. Between d1 d2 will be maximized; d3: maximal distance in order use this. Between d2 d3 will be shaded. 3. a mode consider key points: points: key points are considered independently; polyline: key points form a polyline; polyline loop: key points form a closed polyline. The GPU program that blends s is performed both in vertex fragment shaders. The vertex shader is used pre-compute values needed in fragment shader where main process is realized. The general form algorithm is: For a given fragment 1. Compute closest distance d key points according mode used. 2. For each Compute v, weighted value be applied depending on distances given by user d. Compute color depending on chosen by user weight it by v. 3. Sum colors previously obtained clamp it between 0 1 for each component. This model has been extended

3 generate aumatically dynamically GPU programs (Boyer 2007). Stwares, like ShaderGen, Rendermonkey, ShaderDesigner allow user write shaders (GLSL) through a convenient GUI. A module was created import directly GPU programs created with se stwares. This process is realized using a parser a generar. At end this process, a new vertex shader a new fragment shader are created, compiled linked form new GPU program. It is able in one pass available new one imported by designer. The new GPU program is n dynamically sent graphic card replaces existing. This process can be repeated while GPU capabilities are available. Generation export GPU program The GPU program created by designer is designed be import in a video game focus player's attention. At previous steps, designer has imported selected se be ed chooses a mode distances according model scene. Then we propose aumatically export this shader. A new shader is n aumatically created where only selected by designer are integrated. This supposes that according mode number key points given by designer only one kind distance computation is achieved (part 1 algorithm). Then we copy functions needed (i.e. each m corresponds one technique) we change number iterations in loop algorithm previously presented (part 2 algorithm). Only parameters (uniform, attribute) needed by GPU program are preserved. Finally vertex fragment shaders are created a description file, written in XML is generated save distances associated each, mode used parameters (uniform, attribute, texture). Then, in video game GPU its description file are imported. The vertex fragment shaders are compiled, linked GPU program is used render scene. Moreover as parameters are specified in file description, y can be modified dynamically change effect. Remark that number key points used can be changed according specifications designer but ir positions can be modified dynamically. IMAGES AND RESULTS In this section we present some images produced by our model. Those have been produced on a PC pentium IV 3.6Ghz with 1Go memory a nvidia graphic card Quadro FX This snapshot comes from a car video game. The user should follow a path in a street. The path is represented by red lines along street. In this example we have texture celshading. The key points are placed at junction two streets path. According distance texture, celshading or a se ones are used render geometry.

4 (textured ). We first present scene rendered with one technique same with ed s. Remark that with same GPU program during video game execution, in first image produced with our model one key point is used in second one two key points are used. The second one can be used for example guide player an objective. Here we present anor one can see on s are ed. Last user is not on right path view. As building, view,! Anor example with an helicopter simulation video game. Here designer chooses 2D (luminance sobel filter) 3D In se cases, frame rate decreases 20% (car) or 10% (helicopter). The model performs a GPU program according GPU program selected by designer. So generally, performance generated GPU program is equivalent slower GPU program imported.

5 on high scale scenes, In IASTED Visualization, Imaging, Image Processing (VIIP) 2007, Palma de Majorque, Spain. Boyer V Sobczyk D. Enhancing information on large scenes by ing s. In International Symposium on Visual Computing (ISVC) 2006, South Lake Tahoe, USA. Hertzmann A. Zorin D. Illustrating smooth surfaces. In Proceedings SIGGRAPH Lake A, Marshall C, Harris M, Blackstein M. Stylized for scalable real-time 3d animation. In Non-Phorealistic Animation Rendering 2000 (NPAR 00), Annecy, France. CONCLUSION In this paper we have presented a method focus player's attention in video games. It is based on a model that in same frame. The GPU program is aumatically generated can be imported in video game. Moreover parameters GPU program can be modified dynamically in video game. REFERENCES Boyer V. Aumatic dynamic generation GPU programs s enhance informations Mantiuk R., Krawczyk G., Myszkowski K., Seidel H-P. Perception-motivated High Dynamic Range Video Encoding. Proceedings. SIGGRAPH '04 (Special issue ACM Transactions on Graphics), Rademacher P. View-dependent geometry. In Proceedings SIGGRAPH Rost R. OpenGL Shading Language. Pearson, Winkenbach G. Salesin D.H. Rendering parametric surfaces in pen ink. In Proceedings SIGGRAPH 1996.