几何操作中重采样问题数据并行实现  

Data Parallel Implementation of Resampling in Geometric Operation

作　　者：王光[1] 

机构地区：[1]西安文理学院,陕西西安710065

出　　处：《现代电子技术》2010年第10期30-32,36,共4页Modern Electronics Technique

基　　金："十五"国防863预研课题:星上信息处理技术(2002AA714022)

摘　　要：几何操作是计算机图形和计算机视觉的基本操作之一。如何快速、实时地实现几何操作,是现代媒体应用中的主要问题。重采样操作是图像处理中非常关键的全局操作,一般是通过串行处理方法予以解决。在SIMD处理元阵列上进行数据并行处理几乎是空白,为了满足对图像进行实时处理的需要,提出几何操作中重采样操作的数据并行实现方法,并对结论进行了论证。通过SIMD PE阵列之间的数据并行传送,实现了重采样,从而使实现重采样的复杂度由存储器数据流访问实现方法的Ο(MN)降低到Ο(M+N),大大提高了处理速度,能更好地满足图像快速实时处理的需要。Geometric operation is one of the basic operations in computer graphics and computer vision.How to implement the high-speed and real-time operations effectively,is a major problem existing in modern media applications.The resampling operation,a key global operation in image processing,is generally implemented by sequential implementation means.There are few reports about the data parallel implementation of geometric operations on SIMD processor array.In order to meet the requirements of real-time processing of images,a method of the data parallel implementation of resampling in geometric operations is put forward,and the results are proved.Through the data parallel transmission between SIMD PE arrays,the resampling problem of geometric operations was solved in data parallel.The implementation of the data parallel reduces the complexity of resampling from Ο（MN） implemented by accessing memory to Ο（M＋N）,and greatly enhances the processing speed of processor.Thereby,it can meet the requirements of real-time processing of images.

关 键 词：几何操作 重采样 数据映射 数据并行 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]