python中opencv库的使用 (一) 图像的二值化

灰太狼 2022-01-14 03:35 167阅读 0赞

图像处理中，二值图像是经常用到的处理图像，下面简单介绍一下opencv中的二值化处理

import cv2 as cv
    import matplotlib.pyplot as plt
    
    image = cv.imread('./klay.jpg')
    gray_image = cv.cvtColor(image, cv.COLOR_BGR2GRAY)
    # 其中127为设定的阈值，255为设定的最大阈值， 0为黑色 255为白色
    # 像素小于阈值127设为0，大于127的为最大阈值255
    ret, bw_image1 = cv.threshold(gray_image, 127, 255, cv.THRESH_BINARY)
    # 像素小于阈值127设为最大阈值255，大于127的为0
    ret, bw_image2 = cv.threshold(gray_image, 127, 255, cv.THRESH_BINARY_INV)
    # 像素小于阈值127的保持不变，大于阈值设为最大阈值255
    ret, bw_image3 = cv.threshold(gray_image, 127, 255, cv.THRESH_TRUNC)
    # 像素值小于阈值设为0， 大于阈值127的保持不变
    ret, bw_image4 = cv.threshold(gray_image, 127, 255, cv.THRESH_TOZERO)
    # 像素值小于阈值的保持不变， 大于阈值127的设为0
    ret, bw_image5 = cv.threshold(gray_image, 127, 255, cv.THRESH_TOZERO_INV)
    # 大津法通过图像灰度信息自动获取最佳阈值
    ret, bw_image6 = cv.threshold(gray_image, 127, 255, cv.THRESH_OTSU)
    titles = ['image', 'gray_image', 'BINARY', 'BINARY_INV', 'TRUNC', 'TOZERO', 'TOZERO_INV', 'THRESH_OTSU']
    images = [image, gray_image, bw_image1, bw_image2, bw_image3, bw_image4, bw_image5, bw_image6]
    for i in range(len(images)):
        plt.subplot(2, 4, i+1), plt.imshow(images[i], 'gray')
        plt.title(titles[i])
        plt.xticks([]), plt.yticks([])
    plt.show()

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTI2MDE2NDc_size_16_color_FFFFFF_t_70][]

opencv自带一种局部自适应阈值分割，个人比较推荐，对于局部细节要求比较高的场景，应用的效果还是非常不错的

import cv2 as cv
    import matplotlib.pyplot as plt
    image = cv.imread('./klay.jpg')
    gray_image = cv.cvtColor(image, cv.COLOR_BGR2GRAY)
    # 255为设定的最大阈值， 制定的自适应算法 ADAPTIVE_THRESH_MEAN_C 或 ADAPTIVE_THRESH_GAUSSIAN_C
    # 15 表示邻域块大小，计算区域阈值，一般为3,5,7,9...
    # 5 计算阈值过程中的需要减去的常量
    bw_image = cv.adaptiveThreshold(gray_image, 255, cv.ADAPTIVE_THRESH_GAUSSIAN_C, cv.THRESH_BINARY, 15, 5) # 局部自适应二值化
    cv.imshow('', bw_image)
    cv.waitKey(0)
    cv.destroyAllWindows()

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTI2MDE2NDc_size_16_color_FFFFFF_t_70 1][]

可以看出自适应阈值算法的细节更加丰富，效果更好

[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTI2MDE2NDc_size_16_color_FFFFFF_t_70]: /images/20220113/9d6d7b0cd3fc417e94ed5489ac5138bf.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTI2MDE2NDc_size_16_color_FFFFFF_t_70 1]: /images/20220113/a16eeb9858bb4d0dbc9fe90121603b04.png