在 python 中使用 opencv 剪切图像的特定部分

Question

I have an image of an IC die and I want to cut out the marking in the center.The marking is always at this specific position above the circle at the bottom left. The idea is to first find the circle position which I already accomplished with the hough circle transformation. Now I want to cut out the part where the marking is. It should ideally be a not a square or rectangle but something more like in the image:

This is a part of my code:

        cimg = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
        circles = cv2.HoughCircles(morph_image, cv2.HOUGH_GRADIENT, 1.3, 20, param1=50, param2=25, minRadius=15,
                                   maxRadius=19)

        if circles is not None:
            circles = np.uint16(np.around(circles))
            for i in circles[0, :]:
                # Zeichne äußeren Kreis
                cv2.circle(cimg, (i[0], i[1]), i[2], (0, 255, 0), 2)
                # Zeichne Kreiszentrum
                cv2.circle(cimg, (i[0], i[1]), 2, (0, 0, 255), 3)
                # Tupel mit x- und y-Koordinaten des Kreiszentrums
                circle_center = (i[0], i[1])
                print('Die Koordinaten des Kreiszentrums lauten: ', circle_center)
                """cv2.imshow('Kreis', cimg)
                cv2.waitKey(0)
                cv2.destroyAllWindows()"""
        else:
            circle_center = None
            print('Kein Kreis gefunden')
            """cv2.imshow('Kein Kreis', cimg)
            cv2.waitKey(0)
            cv2.destroyAllWindows()"""

so my cicle center has the center position of my circle (eg (124, 370) ). How can I cut out this part of the image automatically? Can I somehow crop it out? Ideally I would want to crop the marking out into another image to inspect it separately but the normal cropping approach with marking_img = img[y:y+h, x:x+w] wouldn't work I guess.

EDIT: Here is the original image:

The output should be like the first image and if it is possible something like this:

So in the end I would want to have 2 images: One image with just the die without the marking and one image with just the marking

Answer 1

Here is one way in Python/OpenCV.

• Read the image
• Read the mask (separately created one time from your other image)
• Convert the mask to gray and threshold it to binary, invert it and make it 3 channels
• Get the center of the circle from your own code. (I have just measured it manually)
• Set the expected x,y offsets of the bottom of the region of text from the center of the circle
• Compute the expected top left corner of the mask from the center of the circle, the offsets and the height of the mask image
• Put the mask into black image the size of the input at that location
• Apply the new mask to the image to make the rest of the image black
• Crop out the region of interest from the top left corner and the size of the original mask
• OPTIONALLY, crop the original image
• Save the results

Input image:

Prepared mask image:

import cv2
import numpy as np

# read image
img = cv2.imread('die.jpg')
ht, wd, cc = img.shape

# read mask as grayscale
mask = cv2.imread('die_mask.png', cv2.IMREAD_GRAYSCALE)

# threshold mask and invert
mask = cv2.threshold(mask,0,255,cv2.THRESH_BINARY)[1]
mask = 255 - mask
hh, ww = mask.shape

# make mask 3 channel
mask = cv2.merge([mask,mask,mask])

# set circle center
cx = 62
cy = 336

# offsets from circle center to bottom of region
dx = -20
dy = -27

# compute top left corner of mask using size of mask and center and offsets
left = cx + dx
top = cy + dy - hh

# put mask into black background image
mask2 = np.zeros_like(img)
mask2[top:top+hh, left:left+ww] = mask

# apply mask to image
img_masked = cv2.bitwise_and(img, mask2)

# crop region
img_masked_cropped = img_masked[top:top+hh, left:left+ww]

# ALTERNATE just crop input
img_cropped = img[top:top+hh, left:left+ww]

cv2.imshow('image', img)
cv2.imshow('mask', mask)
cv2.imshow('mask2', mask2)
cv2.imshow('masked image', img_masked)
cv2.imshow('masked cropped image', img_masked_cropped)
cv2.imshow('cropped image', img_cropped)
cv2.waitKey(0)
cv2.destroyAllWindows()

# save results
cv2.imwrite('die_mask_inserted.jpg', mask2)
cv2.imwrite('die_masked_image.jpg', img_masked)
cv2.imwrite('die_masked_cropped.jpg', img_masked_cropped)
cv2.imwrite('die_cropped.jpg', img_cropped)

Mask inserted in black image:

Masked image:

Crop of masked image:

(Optional) Crop of input image: