Abstract:  A hybrid method was proposed for cell image segmentation based on histogram analysis using scale space approach and probability relaxation. Firstly, according to the properties of the histogram scale space of an original image and the result of the multiscale filtering, optimal thresholds were determined for classifying the cells image. Then an iterative probability relaxation operation is applied in order to optimize the coarse segmentation. In the postprocessing step, overlapped cells were spilt. To compare with maximum deviation method and region growing method, some results of segmentation were given. The results show that the proposed method is more effective.

Key words: cell image processing, scale space analysis, probability relaxation, image segmentation, multiscale filtering

摘要： 提出基于直方图尺度空间分析和概率迭代松弛的混合方法分割背景复杂的细胞图像。首先根据原始图像直方图的尺度空间特性和多尺度滤波结果,选取最佳阈值将图像分为多个类。然后,利用迭代的概率松弛法对粗分结果进行优化,并在后处理阶段中切割粘连细胞。将该方法与最大方差法和区域增长法进行比较,通过实例表明该方法的有效性。

关键词:  , 细胞图像处理, 尺度空间分析, 概率松弛, 图像分割, 多尺度滤波