Image Thresholding

Why Color Conversion is important?

In some CV applications, it is necessary to convert color images to grayscale, since only edges and shapes are important

Some applications only require a binary image showing general shapes

The goal of thresholding is to create a simple representation of the image that highlights objects of interest by distinguishing them from the background

Idea of Thresholding

• A threshold “T” is selected
• Any point (x,y) in the image at which f(x,y) > T is called an object point
• The segmented image, denoted by g(x,y) is given by

Thresholding

Syntax

retval, thresholded_image = cv2.threshold(src, thresh, maxval, type)

• src: The source image. This should be a grayscale image.
• thresh: The threshold value. Pixel values above or below this value are processed depending on the thresholding type.
• maxval: The maximum value to use with binary thresholding types. This value is assigned to pixels that meet the threshold condition.
• type: The type of thresholding to be applied. Some common types include:
  • cv2.THRESH_BINARY
  • cv2.THRESH_BINARY_INV
  • cv2.THRESH_TRUNC
  • cv2.THRESH_TOZERO
  • cv2.THRESH_TOZERO_INV
• Returns:
  • retval: The threshold value used (often ignored in simple use cases).
  • thresholded_image: The output image after thresholding.

Types

Binary Thresholding

Condition: If the pixel value is greater than 127 (threshold) set it to (maxVal) 255

cv2.threshold(img,127,255,cv2.THRESH_BINARY)

Binary Thresholding – Inverted

Condition: If the pixel value is greater than 127 (threshold) set it to  0

cv2.threshold(img,127,255,cv2.THRESH_BINARY_INV)

Truncate Thresholding

Condition: If the pixel value is greater than 127 (threshold) set it to  127 (threshold

cv2.threshold(img,127,255,cv2.THRESH_TRUNC)

Threshold to zero

Condition: If the pixel value is greater than 127 (threshold) keep it unchanged

cv2.threshold(img,127,255,cv2.THRESH_TOZERO)

Threshold to zero Inverted

Condition: If the pixel value is greater than 127 (threshold) set it to 0

cv2.threshold(img,127,255,cv2.THRESH_TOZERO_INV)

Ways to choose Threshold

• Global Thresholding
• Adaptive Thresholding

Global Thresholding – iterative way

• Global Thresholding
  • A single threshold value is chosen for the entire image. 

Global Thresholding

1. Initial Threshold Selection
   • A common choice is the mean intensity value of the image
2. Segment the image
   • Divide the image into two groups using the current threshold T
   • Group 1: Pixels with Intensity > T (Foreground)
   • Group 2: Pixels with intensity <= T (Background)
3. Calculate the mean intensities
4. Compute new Threshold
5. Repeat step 2 to 4 until convergence

Once convergence is achieved the final value of T_new is used as global Threshold to binarize the image

Solved problem: Refer handwritten notes

Adaptive Mean Thresholding

thresh_mean = cv2.adaptiveThreshold (
    src=img,                  # Input image (grayscale)
    maxValue=255,             # Maximum value to use with the THRESH_BINARY and THRESH_BINARY_INV thresholding types
    adaptiveMethod=cv2.ADAPTIVE_THRESH_MEAN_C,  # Adaptive method (mean calculation)
    thresholdType=cv2.THRESH_BINARY,  # Thresholding type (binary or inverse binary)
    blockSize=3,              # Size of the neighborhood area (must be odd and greater than 1)
    C=5                       # Constant subtracted from the mean or weighted mean
)

Adaptive Gaussian Thresholding

thresh_gaussian = cv2.adaptiveThreshold(
    src=img,                  # Input image (grayscale)
    maxValue=255,             # Maximum value to use with the THRESH_BINARY and THRESH_BINARY_INV thresholding types
    adaptiveMethod=cv2.ADAPTIVE_THRESH_GAUSSIAN_C,  # Adaptive method (Gaussian-weighted mean)
    thresholdType=cv2.THRESH_BINARY,  # Thresholding type (binary or inverse binary)
    blockSize=3,              # Size of the neighborhood area (must be odd and greater than 1)
    C=5                       # Constant subtracted from the mean or weighted mean
)

Views: 1