矩形非均匀采样算法和对数级坐标变换算法的比较分析  

Comparative analysis on rectangular non-uniform sampling algorithm and log-polar transformation algorithm of image

作　　者：訾方[1] 李言俊[1] 张科[1] 赵大炜[1] 

机构地区：[1]西北工业大学航天学院

出　　处：《计算机应用》2007年第7期1619-1622,1640,共5页journal of Computer Applications

基　　金：国家自然科学基金资助项目(60575013);国家留学基金资助项目(中法博士生学院项目)

摘　　要：针对对数极坐标变换引起的图像周边模糊和图像细节丢失等现象,提出了矩形非均匀采样方法。图像中心采样区的大小可根据实际需要进行选取,对其进行均匀采样,对周边区进行分层均匀采样,随距采样中心距离的增大,采样率逐渐降低。对矩形非均匀采样、经典LPT和变参数LPT三者的运算量进行了比较,总体上说,矩形方法具有最少的运算量,而且可以克服对数极坐标变换图像的扭曲,并便于硬件实现。The image＇s peripheral area is blurry and some details may be lost with log-polar transformation（LPT）. So the rectangular non-uniform sampling algorithm （RNS） was proposed. In the method, the size of central area was adjusted according to actual requirements and the central area was sampled uniformly. The peripheral area was layered and the sampling rate in every layer was uniform. Along with the increase of the distance apart from the center, the sampling rate was gradually reduced. The data quantifies of classical LPT variable parameter LPT （VPLPT） and RNS were compared. In general, the calculating quantity of rectangular algorithm is the least. The distortion of image caused by LPT can be overcome. The algorithm can be easily implemented with hardware.

关 键 词：非均匀采样 矩形 对数极坐标变换 算法 分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]