Real-time Rendering With Deformable Transparent Objects

During the numerous beautiful light phenomena in nature, some phenomena cause by the light transmit inside objects are often super wonderful and surprised. These phenom-ena, like refraction and caustics, are also very difficult to simulate in computer. Limited to the graphics hardware's computing ability and storage capacity, previous researcher have been proposed a variety of different methods to rendering them by compromised with speed and quality. Among these methods, ray tracing techniques, due to its physi-cally correct simulation of light transmission path inside objects, could be used to generate high-quality photorealistic images. However, the ray path tracing and ray intersection are computation heavy, traditional methods can hardly be used to render objects in real-time.In this paper, we firstly explained the physical principles behind these beautiful op-tical phenomena caused by transparent objects. Then we introduced the basic methods currently used for rendering:from in pursuit of quality, the off-line ray tracing methods, to in pursuit of speed, the traditional rasterization pipeline. In the following part, we detailed introduced the mainstream methods for rendering transparent objects, form a real-time perspective. From the most basic transparency rendering, to the image-based two interfaces refraction rendering (TIR), and the rendering methods based on volume data refraction, which achieved a high quality comparable to ray-tracing methods. De-tailed comparisons of these methods are discussed.After a full investigation of existing methods used to render transparent objects, we proposed image-based multi interfaces refraction (MIR) and related effects rendering framework. The framework using a novel method to generate multi maps representation of transparent objects, which is greatly suitable for rasterization rendering pipeline. Then the framework introduced a method to implement the refraction and related light effects by using the multi maps representation. Our application is fully implemented by currently popular rendering API on low-price graphics hardware, and we use many different models to complete the experiment. Our approach is general in the computer games and movies. Our approach is general in the sense that it can be used with any object representation that can be rasterized on a programmable GPU.Compared with existing methods, our multi interfaces based rendering method has high interactive rate, high realistic rendering features.●Compared with the two interfaces refraction (TIR), multi interfaces refraction (MIR)'s rendering result is much closer to the rendering results by using off-line ray tracing rendering. This is because successfully tracing the real path when a ray transport insider a transparent objects with complex structure.●Compared with the volume based refraction path calculation methods, multi inter-face based rendering methods deal with sets of two dimensional data. As a result our method has speed advantages.On the other hand, the proposed framework in this paper has a variety of expanded applications. The multi maps generation method can be used to generate a very useful data representation. It almost can transform any representations which can be used for rasterization. Our multi interfaces refraction (MIR) method can be used to not only render many other refraction related effects, but also to render other difficult effects like sub-surface reflection and total internal reflection. And it can be directly used by some height field based simulation, like fluid animation and some pre-computed application like Imposter.The main innovations in this paper are as follows:●This paper presents a new multi maps generation methods without any pre-process algorithm. This method is based on a single pass voxelization and re-rendering, the first pass is a single pass coarse grid voxelization, and the second pass is accurate de-termine of respective layers by using the grid data. Contrary to previous approaches for interactive refraction at multiple interfaces, the proposed techniques does not require any preprocessing. As a result, it can be directly applied to objects under-going shape deformations, which is a common and important feature for character animation in computer games and movies.●This paper explores the limitation of the two interfaces ray intersection algorithm. We extend the ray intersection algorithm to multi interfaces. We proposed a multi interfaces based adaptive size linear search combined with binary search method. Without any heavy calculation consumption, we greatly solve the problems caused by introduce a multi interfaces representation like discontinuity on a single interfaces and the different distance scales between the continuous layers.●With the above mentioned word, we integrated existing rendering equation, approx-imately achieved the refraction with nearby geometry and reflection, caustics, shad-ows and other effects caused by light transmit inside transparent objects. Compared with the current pipeline, the whole process does not increase in time-consuming, which keeps our rendering in real time.