纹理压缩技术的研究与实现

【作者】 罗冠；

【导师】 王淑敏；

【作者基本信息】 西北工业大学 ， 电路与系统， 2001， 硕士

【摘要】 真实感图形实时绘制技术是虚拟现实、动态仿真、实时可视化等的实现基础，一直是计算机图形学领域的研究重点之一。随着图形应用领域的不断扩大，应用内容的复杂化和应用平台的通用化，传统的真实感图形绘制算法面临越来越大的挑战。寻找基于现有计算机硬件性能的实时真实感绘制算法，已是计算机图形学面临十分紧迫的研究课题。 纹理映射是解决场景真实感和实时性矛盾的重要途径之一。纹理可以在不增加场景中几何对象模型复杂性的前提下得到逼真的表面细节，得到越来越广泛的研究和应用。然而在有限的系统带宽的前提下传输大量的纹理数据到纹理缓存，不仅会造成系统数据传输瓶颈效应，而且还会占用大量宝贵的纹理缓存。基于这样的现状，本文考虑通过纹理压缩技术来解决这一问题。 纹理压缩不仅可以提高纹理缓存的利用效率，使采用更大、更高分辨率的纹理成为可能，而更大、更高分辨率纹理图像的采用，将使得虚拟环境中的景物表面看起来更加真实细腻；更能有效地减轻系统数据传输的负担，避免了由于纹理缓存频繁交换带来的系统性能下降，这样就在一定程度上平衡了真实感图形的实时绘制问题。 论文按照纹理映射技术、图像编码技术和纹理压缩技术的顺序进行组织，从纹理本质特征入手，以纹理映射的基本原理和基于矢量量化的图像编码技术为主要内容进行展开，最后提出并实现了适合于纹理图像压缩编码的混合编码算法。实验证明，混合编码算法无论是对缓变，还是对锐变纹理图像，较之传统的图像编码算法具有固定码率编码和独立纹理查找的优点，同时也比S3TC纹理压缩算法具有更高的压缩比和更好的图像重建品质。由于混合编码算法需要将纹理块进行分类压缩，码本的建立相对来说也比较复杂，因此编码时间较长，这还需要在以后的工作中作进一步的改进。更多还原

【Abstract】 Efficient method of real-time realistic rendering is urgently required by increasing applications, such as virtual reality, dynamic simulation, real-time visualization, etc. With the improvement of scene complexity and the generalization of application platform, the contradiction between the speed and quality of existing rendering algorithms has become more and more protruding. Texture mapping is one of the most important approaches in solving the contradiction between real-time rendering and realistic rendering. Much more realistic surface detail can be got without increasing the complexity of geometric model in the visual scene through it. However, if texture is used without limitation, it will soon use up all the display memory, and may cause transmission bottleneck in the system data bus. Texture compression is undoubtedly a good way to solve this problem. First, texture compression can promote the efficiency of display memory and can manage the texture in a fairly efficient way. Thus more bigger and higher resolution textures can be used in one scene to make it more realistic. On the other hand, because texture compression can reduce the texture data, so there is much less need for the data bus to transfer the texture data to display memory, thus can alleviate the burden of system data bus. This paper is organized on the order of texture mapping, image coding and texture compression. The essence of texture is firstly discussed, then the base principle of texture mapping and the technology of image coding based on vector quantization is analyzed in detail. In the end, a hybrid texture compression algorithm is put forward and realized. The efficiency of this method is verified by experimental results. Some deficiencies of it are also presented and analyzed.更多还原