高级检索

分布式并行绘制系统中几何指令流压缩的研究与实现

Research and Implementation of Geometry Instruction Stream Compression in Distributed Parallel Rendering System

  • 摘要: 对分布式并行绘制系统的几何指令流进行压缩能缓解网络带宽瓶颈.对操作码使用LZW算法,对法向量使用球面对称网格剖分算法,对颜色和位置数据使用DPCM型预测编码算法,根据位置数据的特殊性,使用了4类预测器和自适应量化算法.对几何指令流组合使用多种压缩算法取得了良好的效果,在几何模型质量基本没有损失的情况下,指令平均长度压缩到原来的1/3左右,执行速度达到了400指令/ms.

     

    Abstract: Compression of geometry instruction stream in distributed parallel rendering systems can alleviate network bandwidth bottleneck. LZW method is used in op-codes compression. A sphere symmetric-grid method is used for normal compression. Prediction delta coding with DPCM model is used in position and color compression. 4 kinds of predictions and adaptive quantization are used in position compression. Multiple approaches work well in geometry instruction stream compression. The average length of instructions is compressed to 1/3 of the original one with almost no loss of quality and the executing speed is better than 400 instructions/ms.

     

/

返回文章
返回