细胞分割试玩

老师提供的试玩代码

#Segmentation Function
def Watershed(image, number):
    """
    Perform Otsu Segmentation and Watershed,
    Show Original and Segmented Image side by side
    Return Segmented Objects
    :param image: Image to Segment
    :param number: Image Identifier (Int)
    :return: Segmented Objects
    需要先调成灰度图
    """
    #ski.filters库，自动阈值分割，识别细胞
    threshold = filters.threshold_otsu(image)
    cells = image >= threshold/2
    #Scipy.ndimage库，计算细胞间距离
    distance = ndi.distance_transform_edt(cells)
    #ski.feature库，返回图像中局部峰值坐标，创建mask
    local_max_coords = feature.peak_local_max(distance, min_distance=7)
    local_max_mask = np.zeros(distance.shape, dtype=bool)
    local_max_mask[tuple(local_max_coords.T)] = True
    #ski.measure库进行连通区域标记
    markers = measure.label(local_max_mask)
    #分水岭分割算法
    segmented_cells = segmentation.watershed(-distance, markers, mask=cells)
    #有颜色的标记
    color_labels = color.label2rgb(segmented_cells, image, alpha=0.4, bg_label=0)
    io.imsave('Image_'+str(number)+'.tif', color_labels)
    fig, ax = plt.subplots(ncols=2, figsize=(5, 5))
    ax[0].imshow(image, cmap='gray')
    ax[0].set_title("Image "+str(number))
    ax[1].imshow(color_labels)
    ax[1].set_title("Image "+str(number)+" Segmented")
    plt.show(block=False)
    return segmented_cells

先用ski.filters库 图像自动阈值分割自动指定一个阈值，从而来实现分割。
用Scipy.ndimage库，计算细胞间距离
用ski.feature库，返回图像中局部峰值坐标，创建mask
用ski.measure库进行连通区域标记
用ski.segmentation库的分水岭分割算法对图像，标记，mask进行分割。

结果如下：