自适应的遥感纹理并行压缩解压算法  

Adaptive Remote Sensing Texture Parallel Compression and Decompression

作　　者：陆筱霞[1,2] 李思昆[1] 

机构地区：[1]国防科学技术大学计算机学院,长沙410073 [2]中原工学院计算机学院,郑州450007

出　　处：《计算机辅助设计与图形学学报》2013年第5期599-606,共8页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家"九七三"重点基础研究发展计划项目(2002CB312105);国家"八六三"高技术研究发展计划(2006AA01Z309)

摘　　要：为了降低实时地形绘制任务中遥感纹理图像的装载时间,提出一种基于矢量量化的自适应遥感纹理压缩解压算法.基于人类视觉特性优化设计了量化器的纹理块相似度函数,并根据图像特性设计了基础阈值的自动计算方法,可通过自动调节局部阈值来适应遥感图像纹理复杂、局部差异性大的特点;对码字在逻辑邻域内进行优化处理,提高了码书的质量;对索引信息设计了支持随机访问的压缩方法,保证对压缩后索引信息的随机访问.为提高压缩解压速度,充分考虑图形硬件的并行特性,设计了高效的并行算法.实验结果表明,该算法能够在保证压缩率的基础上形成重构质量较好的压缩数据,并且对不同图像都能取得较好的压缩效率;利用图形硬件大幅提高处理速度,完全满足绘制对纹理解压的实时性要求.To decrease the loading time of remote sensing texture in real-time terrain rendering, an adaptive compression and decompression algorithm based on Vector Quantization is presented. Considering the property of human visual system, the similarity function between image blocks is optimized for the quantizer. A fundamental threshold is designed to be computed automatically, which can adjust local threshold to suit the characteristics of remote sensing image, such as complexity and large local difference etc. The code words are also optimized among logical neighborhood to improve codebook^s quality. For the indexes, a new compression method is designed to assure random access in decompression. At last, the above algorithm is implemented on graphics hardware to accelerate the compression and decompression process. Experimental results indicate that the algorithm can generate the compression data with better reconfiguration quantity. The algorithm can also obtain efficient compression for various resolutions and image types. The performance is significantly increased, which totally satisfies the requirement of real-time rendering.

关 键 词：纹理压缩 量化器 人类视觉特性 自适应阈值 索引压缩 图形硬件并行 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]