Real-time texture sampling and reconstruction with wavelet filters

Abstract

Currently, the use of the 2D wavelet transform in texture compression for real-time texture mapping on the GPU is limited. The main cause of this is the lack of real-time texture filtering implementations which do not require specialized hardware. This work proposes a novel system to perform 2D wavelet reconstruction and bilinear texture filtering using a high performance GPU shader. The system is able to generate a performant GLSL shader for arbitrary wavelet filter configurations. This goes beyond earlier works in the literature proposing Haar wavelet and Discrete Cosine Transform (DCT) implementations on the GPU. We analyse the shader performance and run-time complexity for several wavelet filters. The experimental results show that filters longer than Haar are deployable on the GPU while maintaining accurate texture filtering and real-time performance. I. INTRODUCTIONTexture compression has been an essential tool for real time rendering for the past fifteen years. The DXT family of codecs [1] are the most popular texture compression systems, all relying on block truncation coding techniques. Despite their popularity, the attainable compression performance obtained with such simple block-based quantization techniques is modest. In order to further improve compression performance, recent works focus on transform-based texture compression systems, mostly based on the DCT transform [2] and the Haar wavelet transform [3], [4], [5].The transform decorrelates the input signal prior to quantization, which is known to improve the compression gain relative to quantization alone, as performed by the codecs in the DXT family. Moreover, transform-based codecs allow for more freedom than hard-coded DXT codecs. The decomposition of an image into subbands enables features such as scalable compression, per subband bitrate allocation and inherent mip-mapping support.Most of the work in the area of transform-based texture compression cannot be deployed in practical applications, as texture filtering -an essential part of texture mapping -has some serious performance drawbacks or requires a hardware implementation. Additionally, the wavelet based systems are all limited to the Haar transform, which is known to be