尿沉渣检测中的一种自动图像分割算法

An Automatic Algorithm for Image Segmentation in Urine Sediment Examination

目的研究一种自动图像分割算法在尿沉渣检测中的应用。方法将水平集结合Mumford-Shah模型的算法应用于尿沉渣检测,改进章毓晋的图像分割评价指标,并结合计算时间对分割算法进行定量实验评价。结果实验结果表明,与区域生长、分水岭、期望值最大法以及窄带水平集等已有方法比较,本文方法对尿沉渣图像的分割没有过分分割现象,在噪声干扰的情况下具有较高的分割精度,而且耗时少,仅需5.42s。结论评价结果表明本文的算法能够自动分割尿沉渣图像中的细胞,并具有分割精度高,速度快,对噪声鲁棒性强的特点。

Objective To study an accurate algorithm for automatic image segmentation in urine sediment examination. Methods The Mumford-Shah model and level set method were integrated and used to segment the urine sediment image. The algorithm was evaluated by simulation and real data experiment with the improved version of Zhang＇s criterion. Results The Mumford-Shah model based Level Set algorithm could eliminate the over-segment produced by the Level Set, and always had a lowest as compared with the other three algorithms, such as expectation maximization （EM）, region grow and watershed. Timing results showed that the narrow band Level Set algorithm had a highest computational expense （ 〉 1. 8 × 10^4 s） while the Mumford-Shah model based Level Set algorithm was much faster （5.42 s）. Conclusion The Mumford-Shah model based Level Set algorithm can achieve urine sediment examinations accurately with both fast speed and strong robustness to the noise.