numpy.maximum.reduce 不返回最大值

Question

您好，我正在尝试查找已使用 numpy.maximum.reduce 的二进制掩码的最大值。我有一组二进制掩码，以便从所有掩码中找到最大值（根据我的说法，它指向图像中的最多边缘）并避免重叠。因此我使用numpy.maximum.reduce 来找到所有二进制像素的最大值。但结果并不如预期。它没有将白色区域显示为最大值，并将灰色区域作为最终输出掩码（最大值）。为了得到面具，我对边缘检测到的图像进行了膨胀和模糊。那么这会影响最终结果吗（意味着比较掩码以找到最多的边缘）？

Q1.找到最大值并避免重叠区域是否正确？

Q2.它会返回边缘最多的区域吗？

Q3.我应该使用蒙版还是原始边缘检测图像来获得最多边缘？

Results of combining masks Individual Masked Images

for file in glob.glob("images/*.jpg")):
        img = cv2.imread(file)  


        #edge detection
        canny = auto_canny(img)


        #Dilation(Morphological function to increase edge width)
        img_dilate = cv2.dilate(canny, (3,3), iterations = 1)


        #Gaussian Blur to blur the edges to remove noise
        mask= ndimage.gaussian_filter(img_dilate, sigma=5)
        mask[mask< 30] = 0
        mask[(mask>= 30) & (mask< 70)] = 30
        mask[(mask>= 70) & (mask< 110)] = 110
        mask[mask> 110] = 255



#       Retrieve regions from original images
        res = cv2.bitwise_and(img, img, mask = mask)
        mask_inv= cv2.bitwise_not(mask)


        #adding the extracted regions to background image
        main_ = cv2.bitwise_and(im_1, im_1, mask = mask_inv)
        result = cv2.add(main_, res)

        cv2.imshow("result",mask)
        cv2.waitKey(0)

        image_1.append(mask) #list of masks
        image_2.append(res) #list of all extracted regions


for i in range(0,len(image_1)):
    max_ = np.maximum.reduce([image_1[i]])
cv2.imshow("max",max_)
cv2.waitKey(0)

更新

max_img = np.zeros((image_1[0].shape[:3]),dtype=np.uint8)
max_img = np.maximum(max_img,image_1)
cv2.imshow("max",max_img)
cv2.waitKey(0)

使用此更新部分，我遇到了error: (-206) Unrecognized or unsupported array type in function cvGetMat

Answer 1

这是一个使用 Python/OpenCV/Numpy 逐像素获取 4 个二进制掩码图像的最大值的简单示例

4 个面具：

import cv2
import numpy as np

# read masks
mask1 = cv2.imread('mask1.png')
mask2 = cv2.imread('mask2.png')
mask3 = cv2.imread('mask3.png')
mask4 = cv2.imread('mask4.png')

# check shape so that image of zeros (below) is created compatible
print(mask1.shape)

# make list
masks = [mask1, mask2, mask3, mask4]

# initialize maximum array to zeros
image_max = np.full((mask1.shape), (0,0,0), dtype=np.uint8)

# get maximum pair-wise from mask and previous maximum image
for mask in masks:
    image_max = np.maximum(image_max, mask)

# show results
cv2.imshow('maximum', image_max)
cv2.waitKey(0)
cv2.destroyAllWindows()

# save results
cv2.imwrite('mask1-4max.png', image_max)