在python中创建圆圈以遮盖图像并计算每个圆圈中的像素

Question

在python中，我试图将图像分成多个圆圈并计算每个圆圈中黑色像素的数量。

例如，我有一个用鱼眼镜头拍摄的图像（半球形图像）（如下图所示），我想将该图像划分为小圆圈，以捕获图像的一部分，从中间的小圆圈到整个图片。  

我想将图像分成x圈，每次捕获部分图像（请参见下面的图像）

一旦我有了圆形图像，我就可以计算每个图像中的像素数。

我尝试过：     Image=Image.new("RGB", (2000,2000)) draw = ImageDraw.Draw(image) draw.ellipse((20,20,1800,1800),fill(255,255,255)

然后从中创建一个蒙版，但是无论我如何在绘制中更改数字。椭圆形圆圈只会捕获整个图像，但会使图像本身变小。

任何有关解决此问题的想法或建议，将不胜感激！

Answer 1

您应该在OpenCV中查看此类任务。您可以将圆转换为整个轮廓并计算圆的半径。然后，您可以绘制圆形并将其绘制在蒙版上，然后执行cv2.bitwise_and在图像上绘制圆形ROI。您可以使用自己选择的整数（在我的情况下为10）的ROI圆的半径进行迭代和运算。希望能帮助到你。干杯！

示例代码：

import cv2
import numpy as np

img = cv2.imread('circle.png')
h, w = img.shape[:2]
mask = np.zeros((h, w), np.uint8)

gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
kernel = np.ones((10,10),np.uint8)
opening = cv2.morphologyEx(thresh,cv2.MORPH_CLOSE,kernel, iterations = 2)
_, contours, hierarchy = cv2.findContours(opening,cv2.RETR_TREE,cv2.CHAIN_APPROX_NONE)
cnt = max(contours, key=cv2.contourArea)
extLeft = tuple(cnt[cnt[:, :, 0].argmin()][0])
extRight = tuple(cnt[cnt[:, :, 0].argmax()][0])
radius = (extRight[0] - extLeft[0])/2
print(extRight[0], extLeft[0])
print(radius)

M = cv2.moments(cnt)
cx = int(M['m10']/M['m00'])
cy = int(M['m01']/M['m00'])
print(cx, cy)

for i in range(1,30):
    if i*10<radius:
        print(i*10)
        cv2.circle(mask,(cx,cy), i*10, 255, -1)
        res = cv2.bitwise_and(img, img, mask=mask)
        pixels = np.sum(res == 255)
        cv2.putText(res,'Pixel count: '+str(pixels),(30,30), cv2.FONT_HERSHEY_COMPLEX, 0.5, (255,255,255), 1, cv2.LINE_AA)
        cv2.imshow('img', res)
        cv2.waitKey(0)
        cv2.destroyAllWindows()
    else:
        res = cv2.bitwise_and(img, img, mask=opening)
        pixels = np.sum(res == 255)
        cv2.putText(img,'Pixel count: '+str(pixels),(30,30), cv2.FONT_HERSHEY_COMPLEX, 0.5, (255,255,255), 1, cv2.LINE_AA)
        cv2.imshow('img', res)
        cv2.waitKey(0)
        cv2.destroyAllWindows()
        break

结果：

编辑：

尝试使用其他方法计算中间值

import cv2
import numpy as np

img = cv2.imread('circle.png')
h, w = img.shape[:2]
mask = np.zeros((h, w), np.uint8)

gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
kernel = np.ones((10,10),np.uint8)
opening = cv2.morphologyEx(thresh,cv2.MORPH_CLOSE,kernel, iterations = 2)
_, contours, hierarchy = cv2.findContours(opening,cv2.RETR_TREE,cv2.CHAIN_APPROX_NONE)
cnt = max(contours, key=cv2.contourArea)
cv2.imshow('img22', opening)
extLeft = tuple(cnt[cnt[:, :, 0].argmin()][0])
extRight = tuple(cnt[cnt[:, :, 0].argmax()][0])
radius = (extRight[0] - extLeft[0])/2

x,y,w,h = cv2.boundingRect(cnt)
cv2.rectangle(img,(x,y),(x+w,y+h),(0,255,0),2)

cx = int(x+(w/2))
cy = int(y+h/2)

for i in range(1,30):
    if i*10<radius:
        print(i*10)
        cv2.circle(mask,(cx,cy), i*10, 255, -1)
        res = cv2.bitwise_and(img, img, mask=mask)
        pixels = np.sum(res == 255)
        cv2.putText(res,'Pixel count: '+str(pixels),(30,30), cv2.FONT_HERSHEY_COMPLEX, 0.5, (255,255,255), 1, cv2.LINE_AA)
        cv2.imshow('img', res)
        cv2.waitKey(0)
        cv2.destroyAllWindows()
    else:
        res = cv2.bitwise_and(img, img, mask=opening)
        pixels = np.sum(res == 255)
        cv2.putText(img,'Pixel count: '+str(pixels),(30,30), cv2.FONT_HERSHEY_COMPLEX, 0.5, (255,255,255), 1, cv2.LINE_AA)
        cv2.imshow('img', res)
        cv2.waitKey(0)
        cv2.destroyAllWindows()
        break

编辑2：

好的，所以我从您的第一个示例图像中得出的假设是，您的图像距开始将几乎是一个圆圈。因为不是这样，由于图像非常大，因此您不必以不同的方式计算中心（就像从我的第一次编辑-从边界框算起）并制作更大的内核（40,40）。另外，您必须使i达到范围阈值（例如10000）。这将起作用：

import cv2
import numpy as np

img = cv2.imread('circleroi.jpg')
h, w = img.shape[:2]
mask = np.zeros((h, w), np.uint8)

gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
kernel = np.ones((40,40),np.uint8)
opening = cv2.morphologyEx(thresh,cv2.MORPH_CLOSE,kernel, iterations = 2)

_, contours, hierarchy = cv2.findContours(opening,cv2.RETR_TREE,cv2.CHAIN_APPROX_NONE)
cnt = max(contours, key=cv2.contourArea)
extLeft = tuple(cnt[cnt[:, :, 0].argmin()][0])
extRight = tuple(cnt[cnt[:, :, 0].argmax()][0])
radius = (extRight[0] - extLeft[0])/2

x,y,w,h = cv2.boundingRect(cnt)
cv2.rectangle(img,(x,y),(x+w,y+h),(0,255,0),2)

cx = int(x+(w/2))
cy = int(y+h/2)

for i in range(1,10000):
    if i*10<radius:
        cv2.circle(mask,(cx,cy), i*10, 255, -1)
        res = cv2.bitwise_and(img, img, mask=mask)
        pixels = np.sum(res == 255)
        cv2.putText(res,'Pixel count: '+str(pixels),(30,30), cv2.FONT_HERSHEY_COMPLEX, 0.5, (255,255,255), 1, cv2.LINE_AA)
        cv2.imshow('img', res)
        cv2.waitKey(0)
        cv2.destroyAllWindows()
    else:
        res = cv2.bitwise_and(img, img, mask=opening)
        pixels = np.sum(res == 255)
        cv2.putText(img,'Pixel count: '+str(pixels),(30,30), cv2.FONT_HERSHEY_COMPLEX, 0.5, (255,255,255), 1, cv2.LINE_AA)
        cv2.imshow('img', res)
        cv2.waitKey(0)
        cv2.destroyAllWindows()
        break

Answer 2

如果您对库足够熟悉，那么在纯numpy中做起来最简单：

in_loop = False
for i in range(1, x + 1):
    in_loop = True
    print(i)

if not in_loop:
    raise Exception()

要查看各种数组是什么样的：

# Create some fake data
np.random.seed(100)
fake_im_arr = np.random.randint(low=0, high=2, size=(2000,2000))

# Function definition for creating masks
def create_circle_mask(X_arr, Y_arr, center, radius):
    c_x, c_y = center
    dists_sqrd = (X_arr - c_x)**2 + (Y_arr - c_y)**2
    return dists_sqrd <= radius**2

# Using the two together:
center, radius = (1000, 1000), 5
size_x, size_y = fake_im_arr.shape
mask = create_circle_mask(*np.ogrid[0:size_x, 0:size_y], center=center, radius=radius)
n_black_in_circle = ((fake_im_arr == 1) & mask).sum() # This is your answer (39 in this case)