基于最小误判准则的LED管芯数字图像分割

利用基于模式识别决策理论的最小误判准则，给出了求解最佳分割灰度门限的方法，并以磷化镓衬底LED管芯芯片为算例，利用UMA测度对最小误判准则、双峰准则、最大类间距准则等三种算法在分割质量上的差异进行了定量地分析和比较。实验结果表明，就此类图像而言，最小误判准则在计数的准确性和对原始图像质量的鲁棒性方面均优于其它两种方法。

Digital Image Segmentation of Led''''s Chips Based on Minimum Error Criterion

An algorithm for the optimal grey threshold based on minimum error thresholding criterion in pattern recognition decision theory is given. By UMA measure, minimum error criterion is quantitatively analysed and compared with the bi-modal histogram method and the maximum class separation distance criterion in segmentation quality. Experiments on the chips that are lined with gallium phosphate show that, minimum error thresholding criterion is obviously superior to the other two in the segment quality and the rosbustness.