计算机辅助设计与图形学学报
計算機輔助設計與圖形學學報
계산궤보조설계여도형학학보
Journal of Computer-Aided Design & Computer Graphics
2015年
11期
2087-2093
,共7页
王海霞%蔡逸飞%蒋莉%梁荣华
王海霞%蔡逸飛%蔣莉%樑榮華
왕해하%채일비%장리%량영화
图像处理%密度导向%条纹正则化%光学干涉测量
圖像處理%密度導嚮%條紋正則化%光學榦涉測量
도상처리%밀도도향%조문정칙화%광학간섭측량
image processing%density guided%fringe pattern normalization%optical interferometric techniques
在光学干涉测量技术中, 准确地实现条纹正则化是提取条纹图中的相位信息的前提. 文中基于双重正交带通滤波器的正则优化法, 提出了以条纹密度信息为质量评估标准来引导条纹正则化的优化方法, 包含条纹图局部优化与整体优化两部分, 是对双重正交带通滤波器的正则化方法的改进. 首先以条纹背景强度和振幅具有的局部线性与连续性特征设计能量函数, 并通过最小化能量函数获取背景强度和条纹振幅的值; 然后根据高低密度区域条纹的分析, 提出密度对正则化过程的导向作用以及局部优化的可行性.实验结果表明, 以密度为导向的条纹正则优化法具有均方误差低、抗噪性高等优点; 提出的条纹正则优化方法在操作上不需要对噪声严重的条纹图进行除噪预处理, 对背景强度与振幅的线性与非线性变化可同时处理, 具有较高的条纹处理效率和准确性, 可应用于复杂的条纹图正则化处理.
在光學榦涉測量技術中, 準確地實現條紋正則化是提取條紋圖中的相位信息的前提. 文中基于雙重正交帶通濾波器的正則優化法, 提齣瞭以條紋密度信息為質量評估標準來引導條紋正則化的優化方法, 包含條紋圖跼部優化與整體優化兩部分, 是對雙重正交帶通濾波器的正則化方法的改進. 首先以條紋揹景彊度和振幅具有的跼部線性與連續性特徵設計能量函數, 併通過最小化能量函數穫取揹景彊度和條紋振幅的值; 然後根據高低密度區域條紋的分析, 提齣密度對正則化過程的導嚮作用以及跼部優化的可行性.實驗結果錶明, 以密度為導嚮的條紋正則優化法具有均方誤差低、抗譟性高等優點; 提齣的條紋正則優化方法在操作上不需要對譟聲嚴重的條紋圖進行除譟預處理, 對揹景彊度與振幅的線性與非線性變化可同時處理, 具有較高的條紋處理效率和準確性, 可應用于複雜的條紋圖正則化處理.
재광학간섭측량기술중, 준학지실현조문정칙화시제취조문도중적상위신식적전제. 문중기우쌍중정교대통려파기적정칙우화법, 제출료이조문밀도신식위질량평고표준래인도조문정칙화적우화방법, 포함조문도국부우화여정체우화량부분, 시대쌍중정교대통려파기적정칙화방법적개진. 수선이조문배경강도화진폭구유적국부선성여련속성특정설계능량함수, 병통과최소화능량함수획취배경강도화조문진폭적치; 연후근거고저밀도구역조문적분석, 제출밀도대정칙화과정적도향작용이급국부우화적가행성.실험결과표명, 이밀도위도향적조문정칙우화법구유균방오차저、항조성고등우점; 제출적조문정칙우화방법재조작상불수요대조성엄중적조문도진행제조예처리, 대배경강도여진폭적선성여비선성변화가동시처리, 구유교고적조문처리효솔화준학성, 가응용우복잡적조문도정칙화처리.
Fringe pattern normalization is an essential preprocessing step for the data analysis of the optical inter-ferometric techniques. A novel density guided fringe pattern normalization with optimization method is proposed based on the two orthogonal band pass filter, which includes full fringe pattern normalization and partial fringe pattern normalization. In the proposed method, energy function is designed with linear assumption for both back-ground and amplitude of the fringe pattern. Density information is estimated to guide the normalization process since better normalization performance is usually achieved in high density regions. Further, partial normalization is realized with fringe pattern separation using density information. Both simulation and experiment results vali-date the proposed method to be accurate and robust to noise. The proposed method can deal with fringe pattern with severe noise and complex intensity variation without preprocessing, thus is applicable to various types of fringe patterns.