数字图像的概念与性质

import numpy as np

def distanceTransform(img):

    # 默认传入参数为二值图，子集为0，其余为无穷大(255)

    result = img.copy()

    rows, cols = result.shape

    AL = np.array([[-1, 0], [-1, -1], [0, -1], [1, -1]])

    BR = np.array([[1, 0], [1, 1], [0, 1], [-1, 1]])

    for col in range(cols):

        for row in range(rows):

            for next in AL:

                [next_row, next_col] = [row, col] + next

                if 0 <= next_row < rows and 0 <= next_col < cols:

                    if result[next_row, next_col]+abs(row-next_row)+abs(col-next_col) < result[row, col]:

                        result[row, col] = result[next_row, next_col]+abs(row-next_row)+abs(col-next_col)

    for col in range(cols-1, -1, -1):

        for row in range(rows-1, -1, -1):

            for next in BR:

                [next_row, next_col] = [row, col] + next

                if 0 <= next_row < rows and 0 <= next_col < cols:

                    if result[next_row, next_col]+abs(row-next_row)+abs(col-next_col) < result[row, col]:

                        result[row, col] = result[next_row, next_col]+abs(row-next_row)+abs(col-next_col)

    return result

边缘与边界

边缘：像素与其直接邻接的局部性质，是图像上亮度的不连续点，是矢量。方向与梯度垂直，大小与梯度相等。

边界是与区域有关的全局概念

噪声

加性噪声：噪声与图像信号无关

乘性噪声：噪声幅值与信号本身幅值有关

def hist_equal(img):

    rows, cols = img.shape

    S = rows * cols * 1.0

    hist = cv2.calcHist([img], [0], None, [256], [0, 256])

    cum_hist = np.cumsum(hist)

    T = np.zeros(256)

    out = np.zeros_like(img)

    for i in range(256):

        T[i] = np.round(255.0/S * cum_hist[i])

    for row in range(rows):

        for col in range(cols):

            out[row, col] = T[img[row, col]]

    out = cv2.convertScaleAbs(out)

    return out

 def Robert_detection(img, threshold=-1):

    kernel1 = np.array([[-1, 0], [0, 1]])

    kernel2 = np.array([[0, -1], [1, 0]])

    img1 = cv2.filter2D(img, cv2.CV_32F, kernel1)   # 获取-45°方向的偏导

    img2 = cv2.filter2D(img, cv2.CV_32F, kernel2)   # 获取-135°方向的偏导

    mag = cv2.magnitude(img1, img2)

    mag = cv2.convertScaleAbs(mag)

    if threshold == -1:                             # 如果没有给定阈值则通过Otsu二值化

        _, mag = cv2.threshold(mag, 0, 255, cv2.THRESH_OTSU)

    else:

        mag[mag >= threshold] = 255

        mag[mag < threshold] = 0

    return mag

def Sobel_detection(img, threshold=-1):

    dx = cv2.Sobel(img, cv2.CV_32F, 1, 0)           # 获取水平方向的偏导

    dy = cv2.Sobel(img, cv2.CV_32F, 0, 1)           # 获取竖直方向的偏导

    mag = cv2.magnitude(dx, dy)

    mag = cv2.convertScaleAbs(mag)

    if threshold == -1:

        _, mag = cv2.threshold(mag, 0, 255, cv2.THRESH_OTSU)

    else:

        mag[mag >= threshold] = 255

        mag[mag < threshold] = 0

    return mag

def Prewitt_detection(img, threshold=-1):

    kernel1 = np.array([[-1, -1, -1], [0, 0, 0], [1, 1, 1]])

    kernel2 = np.array([[-1, 0, 1], [-1, 0, 1], [-1, 0, 1]])

    dx = cv2.filter2D(img, cv2.CV_32F, kernel2)

    dy = cv2.filter2D(img, cv2.CV_32F, kernel1)

    mag = cv2.magnitude(dx, dy)

    mag = cv2.convertScaleAbs(mag)

    if threshold == -1:

        _, mag = cv2.threshold(mag, 0, 255, cv2.THRESH_OTSU)

    else:

        mag[mag >= threshold] = 255

        mag[mag < threshold] = 0

    return mag

def Laplacian_detection(img, threshold=-1):

    out = cv2.Laplacian(img, cv2.CV_32F, ksize=3)

    out = cv2.convertScaleAbs(out)

    if threshold == -1:

        _, out = cv2.threshold(out, 0, 255, cv2.THRESH_OTSU)

    else:

        out[out >= threshold] = 255

        out[out < threshold] = 0

    return out