OpenCV中的常用API调用示例（像素操作、阈值处理、添加噪音、滤波、基本形态学函数、边缘检测、轮廓检测、直方图）

有关API的调用与解释、以及代码的理解已写在注释中

像素操作

import cv2 as cv


def show_add(src1,src2):
    dst = cv.add(src1, src2)  # cv.add方法
    cv.imshow('add',dst)


def show_subtract(src1,src2):
    dst = cv.subtract(src1, src2)  # cv.add方法
    cv.imshow('subtract',dst)


def show_divide(src1,src2):
    dst = cv.divide(src1, src2)  # cv.add方法
    cv.imshow('divide',dst)


def show_multiply(src1,src2):
    dst = cv.multiply(src1, src2)  # cv.add方法
    cv.imshow('multiply',dst)


src1 = cv.imread('E:\OpenCV\opencv\sources\samples\data\WindowsLogo.jpg')
src2 = cv.imread('E:\OpenCV\opencv\sources\samples\data\LinuxLogo.jpg')
cv.imshow('src1',src1)
cv.imshow('src2',src2)
show_add(src1,src2)
show_subtract(src2,src1)  # 参数顺序影响结果
show_divide(src1,src2)  # 使用很少。参数顺序影响结果
show_multiply(src1,src2)
cv.waitKey()

 

阈值处理 

import cv2 as cv


src = cv.imread('E:\Chidanta.jpg', 0)  #参数0代表读取灰度图
ret, thresh1 = cv.threshold(src, 127, 255, cv.THRESH_BINARY)  # 二值化阈值处理
ret, thresh2 = cv.threshold(src, 127, 255, cv.THRESH_BINARY_INV)
ret, thresh3 = cv.threshold(src, 127, 255, cv.THRESH_TRUNC)  # truncate(截取),大于阈值部分转为阈值
ret, thresh4 = cv.threshold(src, 127, 255, cv.THRESH_TOZERO)  # 小于阈值部分转为0
ret, thresh5 = cv.threshold(src, 127, 255, cv.THRESH_TOZERO_INV)
cv.imshow('src',src)
images = [thresh1, thresh2, thresh3, thresh4, thresh5]
titles = ['binary', 'binary_inv', 'trunc', 'tozero', 'tozero_inv']
for i in range(5):
    cv.imshow(titles[i],images[i])
cv.waitKey()

 

 添加噪音

import cv2 as cv
import numpy as np


def clamp(pv):
    """防止溢出"""
    if pv > 255:
        return 255
    elif pv < 0:
        return 0
    else:
        return pv


def gaussian_noise_demo(image):
    """添加高斯噪声"""
    h, w, c = image.shape
    for row in range(0, h):
        for col in range(0, w):
            s = np.random.normal(0, 15, 3)  # 产生随机数，每次产生三个
            b = image[row, col, 0]  # blue
            g = image[row, col, 1]  # green
            r = image[row, col, 2]  # red
            image[row, col, 0] = clamp(b + s[0])
            image[row, col, 1] = clamp(g + s[1])
            image[row, col, 2] = clamp(r + s[2])

滤波 

import cv2 as cv
import numpy as np


src = cv.imread('E:\Chidanta.jpg', 0)

# 均值滤波，中心值是周围点的均值
blur = cv.blur(src,(3,3))  # ksize用小括号表示大小，而并不是一个列表。blur：模糊

# 方框滤波，在选择归一化=True时结果与均值滤波相同
box = cv.boxFilter(src, -1, (3,3), normalize=True)

# 高斯滤波。实际上是改变了权重矩阵(卷积核)，利用正态分布函数给出不同的权重
gauss = cv.GaussianBlur(src,(3,3), 1)

# 中值滤波。对于某元素的卷积核，按照大小排序，并取中值代替原值
median = cv.medianBlur(src , 5)

# 将图片横向拼接
dst = np.hstack((blur, box, gauss, median))

cv.imshow('show', dst)
cv.waitKey()

 

基本形态学函数 

import cv2 as cv
import numpy as np
import matplotlib as plt


src = cv.imread('Chidanta.jpg', 0)
kernel = np.ones((5,5), np.uint8)
# 腐蚀
dst1 = cv.erode(src, kernel, iterations=1)
# 膨胀
dst2 = cv.dilate(src, kernel, iterations=1)
# 开运算:先腐蚀后膨胀,去刺儿后还原
opening = cv.morphologyEx(src, cv.MORPH_OPEN, kernel)
# 闭运算:先膨胀后腐蚀
closing = cv.morphologyEx(src, cv.MORPH_OPEN, kernel)
# 梯度运算:膨胀-腐蚀
gradient = cv.morphologyEx(src, cv.MORPH_GRADIENT, kernel)
# 礼帽：原图-腐蚀=刺儿
tophat = cv.morphologyEx(src, cv.MORPH_TOPHAT, kernel)
# 黑帽：膨胀-原图=原图的部分信息
blackhat = cv.morphologyEx(src, cv.MORPH_BLACKHAT, kernel)


images = [dst1, dst2, opening, closing, gradient, tophat, blackhat] # 每个图片是它的一个元素
titles = ['erode', 'dilate', 'open', 'close', 'gradient', 'tophat', 'blackhat']
for i in range(7):
    cv.imshow(titles[i],images[i])
cv.waitKey()
cv.destroyAllWindows()

 

 边缘检测

import cv2 as cv
import numpy as np
import matplotlib as plt


src = cv.imread('Chidanta.jpg', 0)
kernel = np.ones((5,5), np.uint8)

# sobel
sobelx = cv.Sobel(src, cv.CV_64F, 1, 0)  # CV_64F是深度参数，它代表可以取负数
sobelx = cv.convertScaleAbs(sobelx)  # 负数在图像显示时会阶段为0，因此取绝对值可以正确表示所有边界
sobely = cv.Sobel(src, cv.CV_64F, 0, 1)
sobely = cv.convertScaleAbs(sobely)
sobelxy = cv.addWeighted(sobelx, 0.5, sobely, 0.5, 0)
cv.imshow('sobel',sobelxy)

# scharr算子的编程与sobel相同，只是kernel不同
# laplacian 二阶导下的求边缘算法，它不分dx、dy，对噪点十分敏感
laplacian = cv.Laplacian(src, cv.CV_64F)
cv.imshow('laplacian',laplacian)

# canny边缘检测
# 1.使用高斯滤波器平滑图像
# 2.计算梯度与方向
# 3.非极大值抑制（两种方法）
# 4.双阈值，minVal，与maxVal。抑制孤立的弱边缘
canny = cv.Canny(src,50,150)
cv.imshow('Canny',canny)

cv.waitKey()
cv.destroyAllWindows()

 

轮廓检测 

import cv2 as cv
import numpy as np
import matplotlib as plt


src = cv.imread('University.jpg')
srcGray = cv.cvtColor(src, cv.COLOR_BGR2GRAY)
ret, binary = cv.threshold(srcGray, 127, 255, cv.THRESH_BINARY)  # 返回两个参数，否则会出问题
# 新版findContours返回两个参数，否则会报错not enough values to unpack
contours, hierarchy = cv.findContours(binary, cv.RETR_TREE, cv.CHAIN_APPROX_NONE)  # hierarchy:等级
ini_src = src.copy()  # 复制一份，防止drawContours覆盖
# 绘图,drawContours可以没有返回值，它直接就在src上画图了
cv.drawContours(src, contours, -1, (0,0,255), 3)

# 轮廓近似
# epsilon = 0.01*cv.arcLength(contours[0], True)
# approx = cv.approxPolyDP(contours, epsilon, True)

# cv.imshow('近似', approx)
cv.imshow('dst', src)
cv.waitKey()
cv.destroyAllWindows()

 

直方图 

import cv2 as cv
import numpy as np
import matplotlib.pyplot as plt


src = cv.imread(r'E:\PyCharm\OpenCV\University.jpg', 0)
plt.hist(src.ravel(),256)  # 可以对图像直接进行求hist
# plt.show()  # 如果一个文档有多个plt.show，则会有多个窗口显示，否则显示在一个窗口上

# 均衡化，
equ = cv.equalizeHist(src)
plt.hist(equ.ravel(),256)
plt.show()  #在此处阻塞

# 自适应均衡化，其实就是将图像分块后均衡化，并在边界处进行线性插值
clahe = cv.createCLAHE(clipLimit=2.0,tileGridSize=(8,8))
# print(type(clahe))  #<class cv2.clahe>
res_clahe = clahe.apply(src)
cv.imshow('clahe', res_clahe)
cv.waitKey()