基于异常检测的尿沉渣图像分割  

URINARY SEDIMENT IMAGE SEGMENTATION BASED ON ANOMOLY DETECTION

作　　者：李悦[1] 嵇启春[1] 

机构地区：[1]西安建筑科技大学,陕西西安727000

出　　处：《计算机应用与软件》2017年第6期212-216,261,共6页Computer Applications and Software

摘　　要：在尿沉渣图像中,由于其样本特性,使得在细胞图像采集时会有大量的杂质。这些杂质形状不规则,颜色不单一,用传统的图像分割算法难以去除。针对这个问题,提出一种基于异常检测的图像分割算法。该方法用形态学的方法对二值图像进行轮廓提取,根据其轮廓进行特征提取并且进行标记,然后用提取的轮廓特征以及标记构建异常检测模型。最终根据该模型对图象进行分割,并且定量地对该模型进行评价。实验结果表明,基于异常检测模型的尿沉渣检测方法能够以较高精度将杂质从细胞图像中分离。In the urine sediment image, due to its sample characteristics, it makes a lot of impurities in the cell image acquisition. These impurities are irregular in shape, and the color is not single, using traditional image segmentation is difficult to remove. Aiming at this problem, an image segmentation algorithm based on anomaly detection is proposed. The algorithm uses the morphological method to extract the binary image contour, according to its contour feature extraction and marking, and an anomaly detection model is constructed using the extracted contour features and the markers. Finally, the image is segmented according to the model, and the model is evaluated quantitatively. The experimental results show that the urine sediment detection method based on the anomaly detection model can separate the impurity from the cell image with high accuracy.

关 键 词：尿沉渣图像 形态学 异常检测 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]