I have merged two lung images using an online merging tool to display a full image.

The problem is that since the background of each image is black I got an unwanted gap between the lungs in the image displayed.

I would like to know if there is a way to remove an area from an image either with code with an algorithm or with an online tool and reduce the gap between the lungs.

Another approach I checked was using OpenCV with Python for a panoramic image stitching, which I will try as a last resort to connect my images.

My desired result:

CodePudding user response：

Here is one way as mention in my comments for Python/OpenCV.

• Read input and convert to grayscale
• Threshold to binary
• Get external contours
• Filter contours to keep only large ones and put bounding box values into list. Also compute the max width and max height from bounding boxes.
• Set desired amount of padding
• Create a black image the size of max width and max height
• Sort the bounding box list by x value
• Get the first item from the list and crop and pad it
• Create a black image the size of max height and desired pad
• Horizontally concatenate with previous padded crop
• Get the second item from the list and crop and pad it
• Horizontally concatenate with previous padded result
• Pad all around as desired
• Save the output

Input:

import cv2
import numpy as np

# load image
img = cv2.imread("lungs_mask.png", cv2.IMREAD_GRAYSCALE)

# threshold
thresh = cv2.threshold(img, 128, 255, cv2.THRESH_BINARY)[1]

# get the largest contour
contours = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
contours = contours[0] if len(contours) == 2 else contours[1]

# get bounding boxes of each contour if area large and put into list
cntr_list=[]
max_ht = 0
max_wd = 0
for cntr in contours:
    area = cv2.contourArea(cntr)
    if area > 10000:
        x,y,w,h = cv2.boundingRect(cntr)
        cntr_list.append([x,y,w,h])
        if h > max_ht:
            max_ht = h
        if w > max_wd:
            max_wd = w
            
# set padded
padding = 25

# create black image of size max_wd,max_ht
black = np.zeros((max_ht,max_wd), dtype=np.uint8)

# sort contours by x value
def takeFirst(elem):
    return elem[0]
cntr_list.sort(key=takeFirst)

# Take first entry in sorted list and crop and pad
item = cntr_list[0]
x = item[0]
y = item[1]
w = item[2]
h = item[3]
crop = thresh[y:y h, x:x w]
result = black[0:max_ht, 0:w]
result[0:h, 0:w] = crop

# create center padding and concatenate
pad_center_img = np.zeros((max_ht,padding), dtype=np.uint8)
result = cv2.hconcat((result, pad_center_img))

# Take second entry in sorted list and crop, pad and concatenate
item = cntr_list[1]
x = item[0]
y = item[1]
w = item[2]
h = item[3]
crop = thresh[y:y h, x:x w]
temp = black[0:max_ht, 0:w]
temp[0:h, 0:w] = crop
result = cv2.hconcat((result, temp))

# Pad all around as desired
result = cv2.copyMakeBorder(result, 25, 25, 25, 25, borderType=cv2.BORDER_CONSTANT, value=(0))

# write result to disk
cv2.imwrite("lungs_mask_cropped.jpg", result)

# display it
cv2.imshow("thresh", thresh)
cv2.imshow("result", result)
cv2.waitKey(0)

Result:

CodePudding user response：

The Concept:

1. Using the

   Notes:

   1. At the line:

   lung_1 = cv2.copyMakeBorder(img[y1: y1   h2, x1: x1   w1], 20, 20, 20, 20, cv2.BORDER_CONSTANT)
   

   notice that we used h2 instead of h1, as if we were to use the h1 we defined, the np.hstack() method would throw an error due to the different heights of the arrays.

   2. The sorted() at

   (x1, y1, w1, h1), (x2, y2, w2, h2) = sorted(map(cv2.boundingRect, contours))
   

   is to sort the x, y, w, h by their x property, so that the lungs are concatenated from left to right.