基于自适应多尺度的血液细胞图像阈值分割方法研究

Study on Image Segmentation for Blood Cells Based on an Adaptive and Multi-Scale Thresholding Approach

提出一个基于自适应多尺度滤波的阈值分割方法 ,对血液细胞图像进行了自动分割 .首先 ,对图像原始直方图进行多尺度滤波 ,根据它的尺度空间图特性 ,确定合适的滤波尺度 ,然后完成对谷点由“粗”到“精”的定位 ,并作为阈值对原始图像进行分割 ;其次 ,利用松弛方法对分割图像作后处理 ;最后 ,将该方法与其他阈值分割方法做了分割性能的比较 。

Based on an adaptive and scale space filter (ASSF), thresholding approach is proposed for automatic segmentation of blood cell images. Apply ASSF to image-histogram, a scale space image is built. According to the properties of the scale space image, a proper scale of filter is obtained. The valleys are gradually located from coarse level to fine level and the location of valleys are applied to the histogram-based segmentation. A relaxation-based postprocessing is employed for the segmented regions. The results show that the proposed algorithm is effective and robust.