opencv常用函数和常用操作

╰+攻爆jí腚メ 2023-02-17 03:38 16阅读 0赞

### 文章目录 ###

*   *   *  一.安装opencv
         *  二.常用函数
         *  三.常用操作

### 一.安装opencv ###

*  python 3.6.3
 *  pip install opencv-python==3.4.1.15
 *  pip install opencv-contrib-python==3.4.1.15

如果觉得下载网速慢的可以换源.在`C:\Users\${XXX用户}\pip`下新建`pip.ini`，输入下面内容替换成清华源。

[global]
    index-url = https://pypi.tuna.tsinghua.edu.cn/simple

### 二.常用函数 ###

*  **cv2.imread(filepath,flags)** 读取图片，默认是三通道(BGR)的彩色图，flags=0读入灰度图
 *  **cv2.VideoCapture** 读取视频
 *  **cv2.imshow(name,img)** 显示图片,需与cv2.waitkey(0)一起使用，不然只显示一瞬间
 *  **cv2.waitkey(timeout)** 显示图片时间，单位为ms,0代表一直显示
 *  **b,g,r=cv2.split(img)** 颜色通道提取,opencv提取格式为BGR(注意不是RBG)
 *  **cv2.merge(b,g,r)** 重新组合颜色通道
 *  **cv2.imwrite(filename,img)** 保存图像
 *  **cv2.copyMakeBorder(img,top, bottom, left, right,borderType)** 填充边界
 *  **cv2.resize(img, (width, height))** 变换图像大小（**注意img.shape的出来的是(height,width)**)
 *  **cv2.addWeighted(img1,α,img2,β,b)** 为图片添加权重, 混合图像 `αX1+βX2+b`
 *  **cv2.blur(img,ksize)** 均值滤波( `矩形中间的值=矩形内的值相加取平均值`)
 *  **cv2.boxFilter(img, ddepth,ksize, normalize=True)** 方框滤波: (`normalize=True等同于均值滤波， normalize=false代表直接求和不取均值，越界后显示为255`)
 *  **cv2.GaussianBlur(img, ksize, sigmaX)** 高斯滤波:(`离得近的权重高，离得远的权重低`）
 *  **cv2.medianBlur(img,ksize)** 中值滤波:（`矩形内的值排序后取中间值`)
 *  **cv2.erode(img, kernel, iterations)** 腐蚀操作
 *  **cv2.dilate(img1, kernel, iterations)** 膨胀操作
 *  **cv2.morphologyEx(img, cv2.MORPH\_OPEN, kernel)** 开运算(`先腐蚀再膨胀`)
 *  **cv2.morphologyEx(img, cv2.MORPH\_CLOSE, kernel)** 闭运算(`先膨胀再腐蚀`)
 *  **cv2.morphologyEx(img, cv2.MORPH\_GRADIENT, kernel)** 梯度运算(`膨胀-腐蚀`)
 *  **cv2.morphologyEx(img,cv2.MORPH\_TOPHAT,kernel)** 顶帽运算(`原始值-开运算`)
 *  **cv2.morphologyEx(img,cv2.MORPH\_BLACKHAT,kernel)** 底帽运算(`闭运算-原始值`)
 *  **cv2.Sobel(img, ddepth, dx, dy,ksize)** Sobel算子
 *  **cv2.Scharr(img, ddepth, dx, dy)** Scharr算子(`能捕捉到更多的细节`)
 *  **cv2.Laplacian(img, ddepth)** Laplacian算子(`噪点影响很大`)
 *  **cv2.Canny(img, threshold1, threshold2)** 边缘检测(`threshold<value<threshold2,范围越大，边缘检测细节越多`
 *  **cv2.pyrUp(img)** 高斯金字塔:向上采样(`放大图像`)
 *  **cv2.pyrDown(img)** 高斯金字塔:向下采样（`缩小图像`)
 *  **img-cv2.pyrUp(cv2.pyrDown(img))** 拉普拉斯金字塔(`原图-(先缩小后放大)`)
 *  **cv2.findContours(img, mode, method)** 找出轮廓
 *  **drawContours(img, contours, contourIdx, color)** 画出轮廓
 *  **cv2.contourArea(contour)** 轮廓面积
 *  **cv2.arcLength(cnt, True)** 轮廓周长 True表示闭合的
 *  **cv2.approxPolyDP(cnt, epsilon, True)** 轮廓近似(`普朗克算法`)
 *  **cv2.matchTemplate(img, templ, method)** 模板匹配
 *  **cv2.rectangle(img, pt1, pt2, color)** 画矩形
 *  **cv2.calcHist(img, channels, mask, histSize, ranges)** 像素直方图统计
 *  **cv2.equalizeHist(img)** 直方图均衡化

### 三.常用操作 ###

*  **只保留某一通道**

#只保留R通道为例
       img[:,:,0]=0  #将B通道全部置为0
       img[:,:,1]=0  #将R通道全部置为0

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70]

*  **裁剪图片**

img[0:200,0:200]

![在这里插入图片描述][20200611232053725.png]

*  **边界填充**

p_top,p_bottom,p_left,p_right=(50,50,50,50)
     wrap=cv2.copyMakeBorder(img,p_top,p_bottom,p_left,p_right,cv2.BORDER_WRAP)

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 1]

*  **混合图像**

img = cv2.imread('1.png')
     img1 = cv2.imread('2.png')
     #!!!注意shape得出来的是(height,width),resize传入的是(width,height)
     h, w = img1.shape[:2]
     #将img的宽高转为img1的宽高
     img = cv2.resize(img, (w, h))
     #αX1+βX2+b
     img2 = cv2.addWeighted(img1, 0.4, img, 0.6, 0)

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 2]

*  **读取视频**

#读取视频
    from cv2 import cv2
    vc=cv2.VideoCapture('1.mp4')
    #检查是否正确打开
    if vc.isOpened:
        open,frame=vc.read()
    else:
        open=False
    while open:
        ret,frame=vc.read()
        if frame is None:
            break
        if ret==True:
            gray=cv2.cvtColor(frame,cv2.COLOR_BGR2GRAY)
            cv2.imshow('1.mp4',gray)
            #按esc退出视频
            if cv2.waitKey(1000) & 0xFF ==27:
                break
    vc.release()
    cv2.destoryAllWindows()

*  **滤波**:减少噪点

# 均值滤波
    # 矩形中间的值=矩形内的值相加取平均值
    blur = cv2.blur(img, (3, 3))
    # 方框滤波
    # normalize=True等同于均值滤波，normalize=false代表直接求和不取均值，越界后显示为255
    # box = cv2.boxFilter(img, -1, (3, 3), normalize=True)
    # 高斯滤波
    # 离得近的权重高，离得远的权重低
    gaussian = cv2.GaussianBlur(img, (3, 3), 1)
    # 中值滤波
    # 取矩形内的中间值
    median = cv2.medianBlur(img, 3)

左起:原图–>均值滤波–>高斯滤波–>中值滤波  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 3]

*  **腐蚀**

kernel = np.ones((3, 3), np.uint8)
    # 腐蚀操作 
    img= cv2.erode(img, kernel, iterations=1)

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 4]

*  **膨胀**

kernel = np.ones((3, 3), np.uint8)
    img = cv2.dilate(img, kernel, 1)

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 5]

*  **梯度处理**

# Sobel算子
    sobelx = cv2.Sobel(img, cv2.CV_64F, 1, 0, ksize=3)
    sobelx = cv2.convertScaleAbs(sobelx)
    sobely = cv2.Sobel(img, cv2.CV_64F, 0, 1, ksize=3)
    sobely = cv2.convertScaleAbs(sobely)
    sobelxy = cv2.addWeighted(sobelx, 0.5, sobely, 0.5, 0)
    
    # Scharr算子
    # 能捕捉到更丰富的细节
    scharrx = cv2.Scharr(img, cv2.CV_64F, 1, 0)
    scharrx = cv2.convertScaleAbs(sobelx)
    scharry = cv2.Scharr(img, cv2.CV_64F, 0, 1)
    scharry = cv2.convertScaleAbs(sobely)
    scharrxy = cv2.addWeighted(scharrx, 0.5, scharry, 0.5, 0)
    
    #Laplacian算子
    #噪点影响很大
    laplacian=cv2.Laplacian(img,cv2.CV_64F)
    laplacian=cv2.convertScaleAbs(laplacian)

左起:原图–>Sobel算子->Scharr算子–>Laplacian算子  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 6]

*  **边缘检测**

img = cv2.imread('1.png', cv2.IMREAD_GRAYSCALE)
    img1 = cv2.Canny(img, 80, 150)
    #阈值越大，边缘检测的细节越多
    img2 = cv2.Canny(img, 20, 200)

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 7]

*  **图像金字塔**

# 高斯金字塔:向上采样(放大图像)
    img1 = cv2.pyrUp(img)
    # 高斯金字塔:向下采样(缩小图像)
    img2 = cv2.pyrDown(img)
    #拉普拉斯金字塔(原图-(先缩小后放大))
    img3 = cv2.pyrUp(cv2.pyrDown(img))
    img.resize(img3.shape)
    img4 = img-img3

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 8]

*  **轮廓检测**

img = cv2.imread('1.png')
    # 转化为灰度图
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    # 使用二值图
    ret, thresh = cv2.threshold(gray, 127, 255, cv2.THRESH_BINARY)
    # 根据灰度二值图找出轮廓
    binary, contours, hierarchy = cv2.findContours(
        thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)
    # 画出轮廓
    # 这里要将原图copy一下，不然直接就在原图上画了
    img1 = img.copy()
    img2 = cv2.drawContours(img1, contours, -1, (0, 0, 255), 2)

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 9]

*  **轮廓近似**

# 轮廓近似
    # 普朗克算法
    epsilon = 0.1*length
    approx = cv2.approxPolyDP(cnt, epsilon, True)
    img3 = img.copy()
    img4 = cv2.drawContours(img3, [approx], -1, (0, 0, 255), 2)

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 10]

*  **模板匹配**

# 以灰度图方式读入
    img = cv2.imread('1.jpg', cv2.COLOR_GRAY2BGR)
    template = cv2.imread('1_template.png', cv2.COLOR_GRAY2BGR)
    h, w = template.shape[:2]
    res = cv2.matchTemplate(img, template, cv2.TM_SQDIFF)
    min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(res)
    img1=img.copy()
    cv2.rectangle(img1, min_loc, (min_loc[0]+w, min_loc[1]+h), 255, 2)

![在这里插入图片描述][20200612013842917.png]  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 11]

*  **像素直方图统计**

import matplotlib.pyplot as plt
    hist = cv2.calcHist([img], [0], None, [256], [0, 256])
    # ravel():将多维矩阵转为一维数组
    plt.hist(img.ravel(), 256)
    plt.show()

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 12]

*  **像素直方图均衡化**

#直方图均衡化
    import matplotlib.pyplot as plt
    equ = cv2.equalizeHist(img)
    plt.hist(equ.ravel(), 256)

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 13]

[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70]: https://img-blog.csdnimg.cn/20200611231137393.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70
[20200611232053725.png]: https://img-blog.csdnimg.cn/20200611232053725.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 1]: https://img-blog.csdnimg.cn/2020061123294047.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 2]: https://img-blog.csdnimg.cn/20200611235608250.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 3]: https://img-blog.csdnimg.cn/20200612002304191.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 4]: https://img-blog.csdnimg.cn/20200612003319109.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 5]: https://img-blog.csdnimg.cn/20200612003520227.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 6]: https://img-blog.csdnimg.cn/20200612005247517.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 7]: https://img-blog.csdnimg.cn/20200612005753957.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 8]: https://img-blog.csdnimg.cn/20200612012214946.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 9]: https://img-blog.csdnimg.cn/20200612013136776.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 10]: https://img-blog.csdnimg.cn/20200612013419996.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70
[20200612013842917.png]: https://img-blog.csdnimg.cn/20200612013842917.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 11]: https://img-blog.csdnimg.cn/202006120139375.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 12]: https://img-blog.csdnimg.cn/20200612014315417.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2_size_16_color_FFFFFF_t_70 13]: https://img-blog.csdnimg.cn/20200612014448849.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzQ1NDUzMjY2,size_16,color_FFFFFF,t_70