python - OpenCV cv2.ellipse 扩展到更难的情况

Question

我有一个掩码(不是二进制，而是 0-255 范围内的值):



我正在使用 cv2.fitEllipse像:

xy_arr = np.argwhere(mask>0)
xy_arr = xy_arr[:,::-1]
(center_x, center_y), (MA, ma), angle = cv2.fitEllipse(xy_arr)
cv2.ellipse(draw_image, (int(center_x), int(center_y)), (int(MA / 2), int(ma / 2)), int(angle), 0, 360, (0, 0, 255))
cv2.circle(draw_image, (int(center_x), int(center_y)), radius=1, color=(0, 0, 255), thickness=1)

结果:



但是，预期的结果:



问题:

是否可以使用掩码的中间值(0-255 范围内的中间值，不仅是二进制掩码)

如何克服左侧尾部的问题，是否可以根据曲率将其切断？

Answer 1

这是使用形态学来减轻 Python/OpenCV 中尾部的一种方法。

阅读输入

转为灰色

阈值

应用形态打开以平滑尾部

获取非零点

获取凸包点

将椭圆拟合到凸包点

在输入

上绘制椭圆

保存结果

输入:

import cv2
import numpy as np

# read image
img = cv2.imread('blob.png')
hh, ww = img.shape[:2]

# convert to grayscale
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)

# threshold to binary and invert
thresh = cv2.threshold(gray, 252, 255, cv2.THRESH_BINARY)[1]

# apply morphology open to smooth out tail
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (9,9))
smoothed = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel)
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (5,5))
smoothed = cv2.morphologyEx(smoothed, cv2.MORPH_DILATE, kernel)

# fit ellipse on smoothed image
points = np.column_stack(np.where(smoothed.transpose() > 0))
hull = cv2.convexHull(points)
((centx,centy), (width,height), angle) = cv2.fitEllipse(hull)

# draw ellipse on input image
result = img.copy()
cv2.ellipse(result, (int(centx),int(centy)), (int(width/2),int(height/2)), angle, 0, 360, (0,0,255), 1)

cv2.imshow('image', img)
cv2.imshow('thresh', thresh)
cv2.imshow('smoothed', smoothed)
cv2.imshow('result', result)
cv2.waitKey(0)
cv2.destroyAllWindows()

# save results
cv2.imwrite('blob_thresh.png', thresh)
cv2.imwrite('blob_smoothed.png', smoothed)
cv2.imwrite('blob_ellipses.png', result)

阈值图像:



形态平滑图像:



生成的椭圆图像: