基于PURE-LET的激光光斑图像快速去噪算法  

PURE-LET-based fast denoising algorithm for laser spot imagery

作　　者：孙继平[1] 江静[1,2] 

机构地区：[1]中国矿业大学北京煤炭资源与安全开采国家重点实验室,北京100083 [2]华北科技学院机电工程系,北京101601

出　　处：《煤炭学报》2013年第9期1710-1714,共5页Journal of China Coal Society

基　　金：国家自然科学基金资助项目(51074169);中央高校基本科研业务费资助项目(JD1201B)

摘　　要：煤矿井下低照度成像环境会导致图像中产生泊松噪声,将引起激光光斑图像的强度和形状分布的不确定性,从而影响激光三角测距的精度。提出了一种基于PURE-LET(Poisson Unbiased Risk Estimator-Linear Expansion of Threshholds)的煤位表面激光光斑图像的快速小波域去噪算法。给出了泊松噪声下小波系数估计MSE(Mean Squared Error)的一个无偏估计子PURE,并将小波系数估计子写作一组基本阈值函数的线性组合以提高算法速度。仿真图像与真实煤仓图像的实验显示,与3种典型图像去噪算法(BayesShrink,Poisson_NLMeans,PURE-LET)相比,提出的PURELET-Smooth算法具有更好的噪声抑制能力,同时具有保持图像边缘结构和快速计算的特点,这在实时光斑质心精确计算和三角测距应用中是一个明显优势。此外此算法具有阈值函数组合系数的快速自动计算特点,适用于自动煤位检测应用。Low illumination imaging environment in coal mine will lead to the Poisson noise contaminating the image- ry, which will cause the uncertainty of spot＇ s intensity and shape, affecting the accuracy of laser triangulation. This pa- per presented a fast denoising algorithm for imagery of laser spot on the surface of coal level based on PURE-LET （Poisson Unbiased Risk Estimator-Linear Expansion of Thresholds）. Under the Poisson noise, we proposed an unbi- ased estimator PURE of the Wavelet coefficients estimator MSE （ Mean Squared Error）. In order to improve the algo- rithm speed, the estimator of wavelet coefficients was expanded as a set linear combination of basic threshold func- tions. Denoising results of simulated images and real-world laser spot images show that the proposed algorithm has bet- ter performance of computation speed and edge-keeping noise suppression, than three other typical image denoising al- gorithm （BayesShrink, Poisson_ NLMeans, PURE-LET）. Additionally this algorithm computes the coefficients of threshold functions automatically, suitable for applications of automatic monitorina of coal levels.

关 键 词：PURE-LET 图像去噪 Harr小波变换 激光光斑 

分 类 号：TD679[矿业工程—矿山机电]