Scribd
Upload
10 views

ImageSegmentation Lung

Uploaded by

Manav Shah
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
10 views

ImageSegmentation Lung

Uploaded by

Manav Shah
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

ImageSegmentationLung

0.0.1 Image Segmentation using OpenCV (Watershed Algorithm)

1. Using watershed Algorithm Steps :

1. Binarization - Find threshold - See Fig. 1.1


2. Sure foreground - erode (removes boundary pixels) - See Fig 1.2
3. Sure background - dilate (increases object boundary to background) - See Fig. 1.3

Basic Idea: Change image to an image, where catchment basins are the objects you want to
segment/identify.

# import necessary modules


import numpy as np
import cv2, matplotlib.pyplot as plt %matplotlib inline

# read image and show


` img = cv2.imread('lung_cancer.jpeg') # in BGR mode
# convert to RGB mode

img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

plt.axis('off') plt.imshow(img)

<matplotlib.image.AxesImage at 0x7fa70299f3c8>

1
Step 1 : Calculating binarization (pixel value is either 0 or 1)

if(pixel_value > threshold):


pixel_value = 1
else:
pixel_value = 0

# performingbinarization # convert to gray scale first

gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY) ret, thresh = cv2.threshold(gray, 0, 255,


cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU) print("Threshold limit: " + str(ret))

plt.axis('off')
plt.imshow(thresh, cmap = 'gray')

Threshold limit: 94.0

<matplotlib.image.AxesImage at 0x7fa7029744e0>

2
# noise removal

kernel = np.ones((3, 3), np.uint8) opening = cv2.morphologyEx(thresh, cv2.MORPH_OPEN,


kernel, iterations = 2)

# sure background area sure_bg = cv2.dilate(opening, kernel,


iterations = 3)

# sure foreground area dist_transform = cv2.distanceTransform(opening,cv2.DIST_L2,5) ret,


sure_fg = cv2.threshold(dist_transform,0.7*dist_transform.max(),255,0)

# Finding unknown region sure_fg =


np.uint8(sure_fg) unknown =
cv2.subtract(sure_bg,sure_fg)

plt.imshow(dist_transform, cmap = 'gray')

<matplotlib.image.AxesImage at 0x7fa7028c1978>

3
fig = plt.figure(figsize = (20, 10)) # to change figsize plt.subplot(131)
plt.imshow(sure_bg, cmap = 'gray') plt.title('Sure background, dilated')

plt.subplot(132) plt.imshow(sure_fg, cmap =


'gray') plt.title('Sure foreground, eroded')

plt.subplot(133) plt.imshow(unknown, cmap = 'gray')


plt.title('Subtracted image, black - sure bg & fg') plt.tight_layout()

# plt.subplots_adjust(wspace = 3)
# fine tuning
# f.subplots_adjust(wspace=3)

4
1. First image above - is dilated image, which increases the object region to background. Whatever
black, is surely background.

2. Second image above - we use distance transform (For connected coins) and then threshold it, to
get the sure coins. (white is sure foreground here).

3. Third image above - we subtract the results in first and second figures to get unknown area. The
white area is unknown area.

Let’s represent sure background and sure foreground with positive integers and unknown area with
0. We use cv2.connectedDots() for this which sets background area to 0, we’ll add one after this to make
sure it does not remain 0 (as we want unknown area to be 0)

ret, markers = cv2.connectedComponents(sure_fg) markers =

markers + 1 markers[unknown==255] = 0

fig = plt.figure(figsize = (20, 10)) # to change figsize plt.subplot(121)


plt.imshow(markers, cmap = 'gray') plt.subplot(122)
plt.imshow(markers)

<matplotlib.image.AxesImage at 0x7fa702bb5e10>

5
markers = cv2.watershed(img, markers) img[markers == -1] = [0, 255,0]

plt.imshow(img)

<matplotlib.image.AxesImage at 0x7fa702bc13c8>

You might also like