在偏置电压较低的情况下，光发射信号较弱，容易被器件的噪声所干扰，造成成像效果不佳．为此，文中建立了光发射显微( PEM) 图像的信号－噪声模型，设计了差分图像定位算法来提高光发射探测的精度．该算法首先通过调整偏置电压获取失效信号强度不同的两组图像，计算出两组图像的差分均值图像，再对其进行均值滤波器、图像二值化等处理，得到统计图像，最后根据统计图像确定集成电路失效点的位置．物理分析结果验证了基于差分图像定位法的集成电路光发射探测技术的准确性．

When the bias voltage is low,PEM (Photon Emission Microscopy) failure signal is weak and is easy tobe influenced by noise,which results in a poor quality of PEM image.In order to solve this problem,a signal- noisemodel of PEM image is established,and a differential image localization algorithm is designed to improve the PEMprecision.First,this algorithm acquires two groups of images with different failure signal strengths by adjusting thebias voltage.Then,the average differential image of these two group images is calculated and dealt with by meansof mean filter and image binarization,thus obtaining a statistical image.Finally,IC failure points can be effectivelylocated according to the statistical image.Physical analysis demonstrates the accuracy of IC PEM based on the pro-posed differential image localization algorithm.