基础:opencv-python-code
图像混合
图片相加
要叠加两张图片,可以用cv2.add()
函数,相加两幅图片的形状(高度/宽度/通道数)必须相同。numpy中可以直接用res = img + img1
相加,但这两者的结果并不相同:
x = np.uint8([250])
y = np.uint8([10])
print(cv2.add(x, y)) # 250+10 = 260 => 255
print(x + y) # 250+10 = 260 % 256 = 4
如果是二值化图片(只有0和255两种值),两者结果是一样的(用numpy的方式更简便一些)。
图像混合
图像混合cv2.addWeighted()
也是一种图片相加的操作,只不过两幅图片的权重不一样,γ相当于一个修正值:
img1 = cv2.imread('lena_small.jpg')
img2 = cv2.imread('opencv-logo-white.png')
res = cv2.addWeighted(img1, 0.6, img2, 0.4, 0)
按位操作
cv2.bitwise_and(), cv2.bitwise_not(), cv2.bitwise_or(), cv2.bitwise_xor()
分别执行按位与/或/非/异或运算。掩膜就是用来对图片进行全局或局部的遮挡。
img1 = cv2.imread('lena.jpg')
img2 = cv2.imread('opencv-logo-white.png')
# 把logo放在左上角,所以我们只关心这一块区域
rows, cols = img2.shape[:2]
roi = img1[:rows, :cols]
# 创建掩膜
img2gray = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY)
ret, mask = cv2.threshold(img2gray, 10, 255, cv2.THRESH_BINARY)
mask_inv = cv2.bitwise_not(mask)
# 保留除logo外的背景
img1_bg = cv2.bitwise_and(roi, roi, mask=mask_inv)
dst = cv2.add(img1_bg, img2) # 进行融合
img1[:rows, :cols] = dst # 融合后放在原图上
平滑图像
均值滤波
img = cv2.imread('lena.jpg')
blur = cv2.blur(img, (3, 3)) # 均值模糊
方框滤波
# 前面的均值滤波也可以用方框滤波实现:normalize=True
blur = cv2.boxFilter(img, -1, (3, 3), normalize=True)
高斯滤波
img = cv2.imread('gaussian_noise.bmp')
# 均值滤波vs高斯滤波
blur = cv2.blur(img, (5, 5)) # 均值滤波
gaussian = cv2.GaussianBlur(img, (5, 5), 1) # 高斯滤波
中值滤波
img = cv2.imread('salt_noise.bmp', 0)
# 均值滤波vs中值滤波
blur = cv2.blur(img, (5, 5)) # 均值滤波
median = cv2.medianBlur(img, 5) # 中值滤波
双边滤波
img = cv2.imread('lena.jpg')
# 双边滤波vs高斯滤波
gau = cv2.GaussianBlur(img, (5, 5), 0) # 高斯滤波
blur = cv2.bilateralFilter(img, 9, 75, 75) # 双边滤波
腐蚀与膨胀
腐蚀
腐蚀的效果是把图片”变瘦”,其原理是在原图的小区域内取局部最小值。因为是二值化图,只有0和255,所以小区域内有一个是0该像素点就为0:
import cv2
import numpy as np
img = cv2.imread('j.bmp', 0)
kernel = np.ones((5, 5), np.uint8)
erosion = cv2.erode(img, kernel)
OpenCV中用cv2.erode()
函数进行腐蚀,只需要指定核的大小就行:
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (5, 5)) # 矩形结构
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (5, 5)) # 椭圆结构
kernel = cv2.getStructuringElement(cv2.MORPH_CROSS, (5, 5)) # 十字形结构
膨胀
膨胀与腐蚀相反,取的是局部最大值,效果是把图片”变胖”:dialation = cv2.dilate(img, kernel)
开/闭运算
先腐蚀后膨胀叫开运算(因为先腐蚀会分开物体,这样容易记住),其作用是:分离物体,消除小区域。这类形态学操作用cv2.morphologyEx()函数实现:
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (5, 5)) # 定义结构元素
img = cv2.imread('j_noise_out.bmp', 0)
opening = cv2.morphologyEx(img, cv2.MORPH_OPEN, kernel) # 开运算
闭运算则相反:先膨胀后腐蚀(先膨胀会使白色的部分扩张,以至于消除/“闭合”物体里面的小黑洞,所以叫闭运算)
img = cv2.imread('j_noise_in.bmp', 0)
closing = cv2.morphologyEx(img, cv2.MORPH_CLOSE, kernel) # 闭运算
形态学梯度
获得物体轮廓
img = cv2.imread('school.bmp', 0)
gradient = cv2.morphologyEx(img, cv2.MORPH_GRADIENT, kernel)
顶帽
原图减去开运算的图tophat = cv2.morphologyEx(img, cv2.MOPRH_TOPHAT, Kernel)
黑帽
闭运算图减去原图blackhat = cv2.morphologyEx(img, cv2.MORPH_BLACKHAT, kernel)
轮廓
cv2.findContours()
&&cv2.drawContours
import cv2
img = cv2.imread('handwriting.jpg')
img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# 使用Otsu自动阈值,注意用的是cv2.THRESH_BINARY_INV
ret, thresh = cv2.threshold(
img_gray, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)
# 寻找轮廓
contours, hierarchy = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
cnt = contours[1]#数组中有0,1两个
cv2.drawContours(img, [cnt], 0, (0, 255, 0), 2)
cv2.imshow('contours', img)
cv2.waitKey(0)
轮廓特征
计算一些轮廓的参数如面积、周长、最小外接矩形等
函数:cv2.contourArea(), cv2.arcLength(), cv2.approxPolyDP()
等
import cv2
import numpy as np
# 载入手写数字图片
img = cv2.imread('handwriting.jpg', 0)
_, thresh = cv2.threshold(img, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)
image, contours, hierarchy = cv2.findContours(thresh, 3, 2)
# 创建出两幅彩色图用于绘制
img_color1 = cv2.cvtColor(image, cv2.COLOR_GRAY2BGR)
img_color2 = np.copy(img_color1)
# 以数字3的轮廓为例
cnt = contours[0]
cv2.drawContours(img_color1, [cnt], 0, (0, 0, 255), 2)
# 1.轮廓面积
area = cv2.contourArea(cnt) # 4386.5
print(area)
# 2.轮廓周长
perimeter = cv2.arcLength(cnt, True) # 585.7716
print(perimeter)
# 3.图像矩
M = cv2.moments(cnt)
print(M)
print(M['m00']) # 同前面的面积:4386.5
cx, cy = M['m10'] / M['m00'], M['m01'] / M['m00'] # 质心
print(cx, cy)
# 4.图像外接矩形和最小外接矩形
x, y, w, h = cv2.boundingRect(cnt) # 外接矩形
cv2.rectangle(img_color1, (x, y), (x + w, y + h), (0, 255, 0), 2)
rect = cv2.minAreaRect(cnt) # 最小外接矩形
box = np.int0(cv2.boxPoints(rect)) # 矩形的四个角点并取整
# 也可以用astype(np.int)取整
cv2.drawContours(img_color1, [box], 0, (255, 0, 0), 2)
cv2.imshow('contours', img_color1)
cv2.waitKey(0)
# 5.最小外接圆
(x, y), radius = cv2.minEnclosingCircle(cnt)
(x, y, radius) = np.int0((x, y, radius))
# 或使用这句话取整:(x, y, radius) = map(int, (x, y, radius))
cv2.circle(img_color2, (x, y), radius, (0, 0, 255), 2)
# 6.拟合椭圆
ellipse = cv2.fitEllipse(cnt)
cv2.ellipse(img_color2, ellipse, (0, 255, 0), 2)
cv2.imshow('contours2', img_color2)
cv2.waitKey(0)
# 7.形状匹配
img = cv2.imread('shapes.jpg', 0)
_, thresh = cv2.threshold(img, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
image, contours, hierarchy = cv2.findContours(thresh, 3, 2)
img_color = cv2.cvtColor(image, cv2.COLOR_GRAY2BGR)
cnt_a, cnt_b, cnt_c = contours[0], contours[1], contours[2]
print(cv2.matchShapes(cnt_b, cnt_b, 1, 0.0)) # 0.0
print(cv2.matchShapes(cnt_b, cnt_c, 1, 0.0)) # 2.17e-05
print(cv2.matchShapes(cnt_b, cnt_a, 1, 0.0)) # 0.418
直方图
直方图简单来说就是图像中每个像素值的个数统计,比如说一副灰度图中像素值为0的有多少个,1的有多少个……直方图是一种分析图片的手段:
使用cv2.calcHist(images, channels, mask, histSize, ranges)计算,其中:
参数1:要计算的原图,以方括号的传入,如:[img]
参数2:类似前面提到的dims,灰度图写[0]就行,彩色图B/G/R分别传入[0]/[1]/[2]
参数3:要计算的区域,计算整幅图的话,写None
参数4:前面提到的bins
参数5:前面提到的range
import cv2
import numpy as np
import matplotlib.pyplot as plt
img = cv2.imread('2.jpg', 0)
hist = cv2.calcHist([img], [0], None, [256], [0, 256]) # 性能:0.025288 s
plt.plot(hist)
plt.show()
cv2.waitKey(0)
自动均衡
#原图与均衡比较
import cv2
import numpy as np
import matplotlib.pyplot as plt
img = cv2.imread('2.jpg', 0)
hist = cv2.calcHist([img], [0], None, [256], [0, 256]) # 性能:0.025288 s
plt.plot(hist)
plt.show()
equ = cv2.equalizeHist(img)
hist2 = cv2.calcHist([equ], [0], None, [256], [0, 256]) # 性能:0.025288 s
plt.plot(hist2)
plt.show()
#v2.imshow('equalization', np.hstack((img, equ))) # 并排显示
cv2.waitKey(0)
模版匹配
用cv2.matchTemplate()
实现模板匹配。首先我们来读入图片和模板:
import cv2
import numpy as np
from matplotlib import pyplot as plt
# 1.模板匹配
img = cv2.imread('1.jpg', 0)
template = cv2.imread('2.jpg', 0)
h, w = template.shape[:2] # rows->h, cols->w
# 6种匹配方法
methods = ['cv2.TM_CCOEFF', 'cv2.TM_CCOEFF_NORMED', 'cv2.TM_CCORR',
'cv2.TM_CCORR_NORMED', 'cv2.TM_SQDIFF', 'cv2.TM_SQDIFF_NORMED']
for meth in methods:
img2 = img.copy()
# 匹配方法的真值
method = eval(meth)
res = cv2.matchTemplate(img, template, method)
min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(res)
# 如果是平方差匹配TM_SQDIFF或归一化平方差匹配TM_SQDIFF_NORMED,取最小值
if method in [cv2.TM_SQDIFF, cv2.TM_SQDIFF_NORMED]:
top_left = min_loc
else:
top_left = max_loc
bottom_right = (top_left[0] + w, top_left[1] + h)
# 画矩形
cv2.rectangle(img2, top_left, bottom_right, 255, 2)
plt.subplot(121), plt.imshow(res, cmap='gray')
plt.xticks([]), plt.yticks([]) # 隐藏坐标轴
plt.subplot(122), plt.imshow(img2, cmap='gray')
plt.xticks([]), plt.yticks([])
plt.suptitle(meth)
plt.show()
同时匹配多个物体
# 1.读入原图和模板
img_rgb = cv2.imread('mario.jpg')
img_gray = cv2.cvtColor(img_rgb, cv2.COLOR_BGR2GRAY)
template = cv2.imread('mario_coin.jpg', 0)
h, w = template.shape[:2]
# 2.标准相关模板匹配
res = cv2.matchTemplate(img_gray, template, cv2.TM_CCOEFF_NORMED)
threshold = 0.8
# 3.这边是Python/Numpy的知识,后面解释
loc = np.where(res >= threshold) # 匹配程度大于%80的坐标y,x
for pt in zip(*loc[::-1]): # *号表示可选参数
right_bottom = (pt[0] + w, pt[1] + h)
cv2.rectangle(img_rgb, pt, right_bottom, (0, 0, 255), 2)
霍夫变换
学习使用霍夫变换识别出图像中的直线和圆cv2.HoughLines()
霍夫直线变换cv2.HoughCircles
霍夫圆变换