如何填充粒子边界缺口以形成连续轮廓？ImageJ/OpenCV方案咨询

方法1：OpenCV 精准闭运算+轮廓筛选+分水岭分割 #

针对缺口填补与粒子粘连的矛盾，可分三步操作：

1. 小核闭运算填补缺口：用小尺寸椭圆结构元素执行闭运算，优先填补小缺口，同时尽量减少粒子粘连

   import cv2
   import numpy as np
   
   img = cv2.imread("your_image.png", 0)
   # 创建贴合粒子轮廓的椭圆结构元素
   kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (3,3))
   closed = cv2.morphologyEx(img, cv2.MORPH_CLOSE, kernel, iterations=1)
   

2. 轮廓筛选去除伪轮廓：提取闭运算后的轮廓，通过面积、轮廓近似筛选出单个粒子的轮廓

   contours, _ = cv2.findContours(closed, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
   filled_img = np.ones_like(img) * 255
   for cnt in contours:
       area = cv2.contourArea(cnt)
       # 根据实际粒子尺寸设置面积阈值
       if 50 < area < 500:
           approx = cv2.approxPolyDP(cnt, 0.02*cv2.arcLength(cnt, True), True)
           cv2.drawContours(filled_img, [approx], -1, 0, thickness=cv2.FILLED)
   

3. 分水岭分割处理轻微粘连：若仍有少量粒子粘连，结合距离变换+分水岭算法分割

   dist_transform = cv2.distanceTransform(closed, cv2.DIST_L2, 5)
   ret, sure_fg = cv2.threshold(dist_transform, 0.3*dist_transform.max(), 255, 0)
   sure_fg = np.uint8(sure_fg)
   unknown = cv2.subtract(closed, sure_fg)
   ret, markers = cv2.connectedComponents(sure_fg)
   markers = markers + 1
   markers[unknown==255] = 0
   markers = cv2.watershed(cv2.cvtColor(img, cv2.COLOR_GRAY2BGR), markers)
   filled_img[markers == -1] = 255
   

方法2：ImageJ 分步修复+精准填充 #

1. 修复边缘缺口：先执行Process > Binary > Close，选择1-2像素迭代次数、椭圆结构元素填补小缺口；若仍有缺口，用Process > Noise > Despeckle去噪后，通过Edit > Selection > Create Selection手动补全大缺口（按住Shift可添加选区）
2. 分割粘连粒子：使用Analyze > Analyze Particles，勾选Exclude on edges和Add to Manager，对粘连粒子用Edit > Selection > Split手动分割，或用Plugins > Segmentation > Watershed自动分割
3. 批量填充：在选区管理器中选中所有粒子选区，执行Edit > Fill填充为黑色

方法3：Scikit-image 图像修复+主动轮廓拟合 #

针对不规则缺口，用图像修复补全边界后再填充：

from skimage import inpaint, io, color
import numpy as np
import cv2

img = io.imread("your_image.png")
gray = color.rgb2gray(img)
# 标记缺口区域（根据实际边缘情况调整）
edges = cv2.Canny(gray, 50, 150)
mask = np.zeros_like(gray)
mask[edges == 0] = 1
# 双调和方程修复图像
restored = inpaint.inpaint_biharmonic(gray, mask, multichannel=False)
# 阈值化后填充
thresh = cv2.threshold(restored, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]
contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
filled_img = np.ones_like(gray) * 255
for cnt in contours:
    cv2.drawContours(filled_img, [cnt], -1, 0, cv2.FILLED)

内容的提问来源于stack exchange，提问作者carol208