基于支持向量机的显微图像聚焦区域选取算法

An algorithm for selecting focus area of micro image based on support vector

聚焦窗口的选取易受噪声、镜头杂质和玻片杂质的影响,本文针对玻片杂质引起聚焦曲线失去理想曲线特性的问题,提出一种基于支持向量机(SVM)的显微图像聚焦区域选取算法。首先,获取当前聚焦平面的图像,通过对比度和邻域相关性系数结合的方法检测是否存在模糊区域,以此作为是否存在玻片杂质的标准;其次,检测出模糊区域后,将当前图片进行色彩模式(CMYK)转化,提取图像的K分量图,对像素点进行处理输出分量图;再次,大致将分量图分为2类,计算子块的梯度和像素值总和进行坐标表示,通过支持向量机的训练,找出最佳分割线,识别并标定杂质区域;最后,剔除玻片杂质区域,重新选取聚焦窗口并进行二次聚焦。实验结果表明,该算法能有效剔除玻片杂质的影响,使得聚焦曲线保持较好的单峰性,且解决了玻片杂质引起焦平面误判的问题,确保选取到内容丰富的子块作为聚焦窗口,提高了聚焦的准确性。

The selection of focusing window is easy to be affected by noise,lens impurity and glass impurity.A support vector machine(SVM)based micro-image focus area selection algorithm is proposed to solve the problem of the loss of ideal curve characteristics caused by glass impurity.Firstly,an image of a current focusing plane is acquired to detect whether a blurred region exists by a method of combining contrast and neighborhood correlation coefficient as a standard for judging whether a glass impurity is present.Secondly,after the blurred region is detected,the current picture is transformed by color mode(CMYK),the K component of the image is extracted and the pixel is processed and the component map is output.Thirdly,the component map is roughly divided into 2 categories.The gradient of the sub-block and the sum of the pixel values are calculated to represent the coordinates.The best segmentation line is found through training of support vector machines,and the impurity regions are identified and calibrated.Finally,the glass impurity region is removed,and the focus window is selected again for secondary focusing.The experimental results show that the algorithm can effectively eliminate the influence of glass impurities,keep the focusing curve unimodal,solve the problem of focal plane misjudgment caused by glass impurities,and ensure that the rich content sub-blocks are selected as the focus window,improving the accuracy of focusing.