Dajing method (Python implementation)

In computer vision and image processing, the Dajing method is used to automatically obtain the threshold of an image, or to convert a grayscale image to a two-valued image. The algorithm assumes that the image contains two categories of pixels (foreground and background pixels) and then calculates an optimal threshold for separating the foreground and background, minimizing the variance between the foreground and background classes.

The following defines the formula for calculating the inter-class variance:
Suppose we use a threshold T to split a grayscale image into a foreground and a background size: Total number of pixels in the image
U: Average grayscale of images
W0: The ratio of the foreground pixel to the full image size u0: The average of the foreground pixel points
W1: The ratio of the background pixel to the whole image size
U0: Average of background pixel points
G: Inter-class variance

u = w0 * u0 + W1 * U1 (1)
g = w0* (u-u0) ^2 + w1* (U-U1) ^2 (2)
Substituting (1) into (2) the following:
g = w0 * W1 * (U0-U1) ^2

Using the traversal method, all thresholds are traversed, and when G is the maximum, the threshold is what we think is the most appropriate threshold value.
Secretly tell you, OpenCV has its own method of thresholding using the Dajing method, as follows:

Use the parameter cv2. After the Thresh_otsu, the Th1 in front is void.

The following is a Python version of the function I wrote myself:

Import NumPy as Npdef otsu_enhance (Img_gray, th_begin=0, th_end=256, th_step=1):    assert Img_gray.ndim = = 2, "Must Inpu T a gary_img "    max_g = 0    suitable_th = 0 for    threshold in xrange (Th_begin, Th_end, th_step):        bin_img = Img_ Gray > Threshold        BIN_IMG_INV = Img_gray <= threshold        Fore_pix = np.sum (bin_img)        Back_pix = Np.sum (bin_ IMG_INV)        If 0 = = Fore_pix: Break        if 0 = = Back_pix:            continue        w0 = float (fore_pix)/img_gray.size< C13/>u0 = Float (np.sum (Img_gray * bin_img))/Fore_pix        w1 = float (back_pix)/img_gray.size        u1 = float (Np.sum (IM G_gray * BIN_IMG_INV))/Back_pix        # intra-class Variance        g = w0 * W1 * (U0-U1) * (U0-U1)        if g > Max_ G:            max_g = g            suitable_th = threshold    return suitable_th

Dajing method (Python implementation)