Installation

python opencv 包的官方网址: https://pypi.org/project/opencv-python/

opencv官网网址 : https://docs.opencv.org/master/

• 直接通过pip安装(有图形界面的PC)

  • 选择1：包含主要的模块: pip install opencv-python
  • 选择2: 全部安装(有些额外的模块) : pip install opencv-contrib-python

• 没有图形界面的服务器上(与👆对应)

  • pip install opencv-python-headless
  • pip install opencv-contrib-python-headless

• 尝试用conda install open-cv,但是找不到包

Common Usage

导入的包为cv2

1. 读取图像

import cv2
cv2.imread("dog.png")

返回值: 像素矩阵[height, width, channel]

类型: numpy.ndarray；

数据类型: unit8;

通道: BGR, 不是RGB; 可以通过src[:,:,::]转换

注意：每个像素的数据类型一定要uint8；否则会出现

error: (-215:Assertion failed) src_depth != CV_16F && src_depth != CV_32S in function ‘convertToShow’

解决办法: np.uint8(image)

2. 显示图像

cv2.imshow("winname", image)

3. 缩放图像

cv2.resize(src, dsize[, dst[, fx[, fy[, interpolation]]]]) --> dst

参数说明:

• src: 读入的图像, [height, width, channel]
• dsize: 变换后的宽和高(width, height) —> (int, int)
• dst: 与src相对应, 即变换完的各点像素矩阵
• fx: x方向(即width方向) 缩放倍数
• fy: y方向(即height方向)缩放倍数
• interpolation: 插值方法
• notes: 转换后的尺寸要么由dsize要么由fx,fy来决定

插值方法	解释如下
INTER_NEAREST	最近邻插值(0)
INTER_LINEAR	双线性插值（默认设置）(1)
INTER_AREA	使用像素区域关系进行重采样（最适合shrink)
INTER_CUBIC	4x4像素邻域的双三次插值 (2)
INTER_LANCZOS4	8x8像素邻域的Lanczos插值(3)

👆标号0,1,2,3代表上采样方法越来越复杂, 图像过度越平滑,质量越好,但是速度越慢

• 如果你已经创建了dst(转换后的尺寸由dst来决定)

  resize(src, dst.size(), dst)

• 修改图片尺寸

  import cv2
  import numpy as np
  
  
  def letterbox_image(img, inp_dim):
      """resize image with unchanged aspect ratio using padding
      img: input image
      inp_dim: 
      """
      img_w, img_h = img.shape[1], img.shape[0] # img:[height, width, channel]
      w, h = inp_dim
      scale = min(w / img_w, h / img_h) 		  # keep aspect ratio
      new_w = int(img_w * scale) 				  # multiply the same number
      new_h = int(img_h * scale)
      # 表示大小时用的是(width,height) --> 返回的是[height, width, channel]
      resized_image = cv2.resize(img, (new_w, new_h), interpolation=cv2.INTER_CUBIC) # cubic
      cv2.imshow("unchanged aspect ratio", resized_image)
      canvas = np.full((h, w, 3), 128)          # create a numpy array having shape of [width, height, c]
  
      # padding with (128,128,128) gray
      canvas[(h - new_h) // 2: (h - new_h) // 2 + new_h, (w - new_w) // 2:(w - new_w) // 2 + new_w, :] = resized_image
  
      return canvas
  
  if __name__ == "__main__":
      filename = "dog-cycle-car.png"
      dog = cv2.imread(filename)
      cv2.imshow("original_dog", dog)
      shrink_dog = cv2.resize(dog, None, fx=0.5, fy=1, interpolation=cv2.INTER_CUBIC)
      dog_keep_ratio_padding = np.uint8(letterbox_image(dog, (dog.shape[0] // 2, dog.shape[0] // 2))) # height/2
      cv2.imshow("dog_keep_ratio_padding", dog_keep_ratio_padding)
      cv2.imshow("shrink_dog", shrink_dog)
      cv2.waitKey(0)
      cv2.destroyAllWindows()

  • 一定要使用cv2.waitKey(0)，否则显示不了图片:（正数为等待的时间,单位毫秒)
  • 使用cv2.destroyWindows(“winname”),可以主动关闭窗口(图片)

  

4. 在图像上画矩形)

cv2.rectangle(img, pt1, pt2, color[, thickness[, lineType[, shift]]]) -> img

The function cv::rectangle draws a rectangle outline or a filled rectangle whose two opposite corners

are pt1 and pt2

pt1,pt2: 两个对角的坐标

color: tuple(R,G,B)

thickness: int 如果为负数,比如-1,那么会填充整个矩形

lineType: 8(默认)：8连接；4：4连接线；CV_AA:锯齿线

注:对于整数坐标的非反锯齿线，使用8连接或4连接的Bresenham算法。粗线以圆角结尾绘制。反锯齿线是用高斯滤波绘制的

shift: 点坐标中的小数位数

5.在图像上写字

cv2.putText(img, text, org, fontFace, fontScale, color[, thickness[, lineType[, bottomLeftOrigin]]])

对于不能渲染的symbol，使用？标记

text: String to be drawn

org: Tuple(x,y) Bottom-left corner of the text string in the image(文本的左下角坐标)

fontFace: 字体

fontScale: 字体大小

color, thickness, lineType: 同👆

bottomLeftOrigin: 当其为真时，图像原点在左下角，否则在左上角(opencv默认为左上)【如果为True，字翻转$180^o$】

enum HersheyFonts {
    FONT_HERSHEY_SIMPLEX        = 0, //!< normal size sans-serif font
    FONT_HERSHEY_PLAIN          = 1, //!< small size sans-serif font
    FONT_HERSHEY_DUPLEX         = 2, //!< normal size sans-serif font (more complex than FONT_HERSHEY_SIMPLEX)
    FONT_HERSHEY_COMPLEX        = 3, //!< normal size serif font
    FONT_HERSHEY_TRIPLEX        = 4, //!< normal size serif font (more complex than FONT_HERSHEY_COMPLEX)
    FONT_HERSHEY_COMPLEX_SMALL  = 5, //!< smaller version of FONT_HERSHEY_COMPLEX
    FONT_HERSHEY_SCRIPT_SIMPLEX = 6, //!< hand-writing style font
    FONT_HERSHEY_SCRIPT_COMPLEX = 7, //!< more complex variant of FONT_HERSHEY_SCRIPT_SIMPLEX
    FONT_ITALIC                 = 16 //!< flag for italic font
};

字体👆

6 示例

import cv2


if __name__ == "__main__":
    filename = "imgs\\dog.jpg"
    dog = cv2.imread(filename)
    cv2.imshow("original_dog", dog)
    c1 = (164, 108)
    c2 = (560, 447)
    color = (255, 0, 0) # red
    cv2.rectangle(dog, c1, c2, color, 2, lineType=cv2.LINE_AA)
    cv2.imshow("rectangle", dog)
    label = "bicycle"
    # (text,fontFace,fontScale,thickness)
    t_size = cv2.getTextSize(label, cv2.FONT_HERSHEY_SIMPLEX, 1, 1)[0] # 
    c2 = c1[0] + t_size[0] + 3, c1[1] - t_size[1] - 4 # c2 at the right-top of original rectangle's top-left corner
    # (width:x, height:y)
    cv2.rectangle(dog, c1, c2, color, -1, lineType=cv2.LINE_AA) # -1 fill the rectangle
    cv2.putText(dog, label, (c1[0], c1[1]-6), cv2.FONT_HERSHEY_SIMPLEX, 1, [225, 255, 255], 1)
    cv2.imshow("rectangle+text", dog)
    cv2.waitKey(0)
    cv2.destroyAllWindows()