import datetime, math, os, sys, logging
import cv2
import numpy as np


# r / b
def ratio_method1(image_path):
    # Load an image
    img = cv2.imread(image_path)
    # img = cv2.imread(image_path, 0)  # Load an image in grayscale

    b, g, r = cv2.split(img)
    cloud = r / b

    ret, cloud_BINARY = cv2.threshold(cloud, 0.6, 255, cv2.THRESH_BINARY)  # 文献中推荐阈值为0.6
    cloudage = cloud_BINARY.sum() / (cloud_BINARY.size * 255)

    print("r/b算法，云量为%f\n" % cloudage)

    tmp = image_path.split('.')
    out = tmp[0] + "_r_b." + tmp[1]
    cv2.imwrite(out, cloud_BINARY)

    # cv2.imshow('image', cloud_BINARY)
    # cv2.waitKey(0)
    # cv2.destroyAllWindows()

    return cloud_BINARY


# (b-r) / (b+r)
def ratio_method2(image_path):
    img = cv2.imread(image_path)

    b, g, r = cv2.split(img)
    cloud = (b-r) / (b+r)

    ret, cloud_BINARY = cv2.threshold(cloud, 0.2, 255, cv2.THRESH_BINARY_INV)
    cloudage = cloud_BINARY.sum() / (cloud_BINARY.size * 255)

    print("(b-r) / (b+r)算法，云量为%f\n" % cloudage)

    tmp = image_path.split('.')
    out = tmp[0] + "_(b-r)_(b+r)." + tmp[1]
    cv2.imwrite(out, cloud_BINARY)

    # cv2.imshow('image', cloud_BINARY)
    # cv2.waitKey(0)
    # cv2.destroyAllWindows()

    return cloud_BINARY


# r / b - (b-r) / (b+r)
def ratio_method1_subtract_ratio_method2(image_path):
    result1 = ratio_method1(image_path)
    result2 = ratio_method2(image_path)

    light_cloud_BINARY = np.zeros_like(result1)

    for x in range(result1.shape[0]):
        for y in range(result1.shape[1]):
            if (result1[x, y] != 0 and result2[x, y] == 0):
                light_cloud_BINARY[x, y] = 255

    tmp = image_path.split('.')
    out = tmp[0] + "_subtract." + tmp[1]
    cv2.imwrite(out, light_cloud_BINARY)

    # cv2.imshow('image', light_cloud_BINARY)
    # cv2.waitKey(0)
    # cv2.destroyAllWindows()


def saturability_method(image_path):
    img = cv2.imread(image_path)
    b, g, r = cv2.split(img)

    # 自己计算饱和度-----> 失败：计算的饱和度不对，是负值
    ones = np.ones_like(b)
    sum = b + g + r
    min_b_g_r = np.zeros_like(b)

    for x in range(b.shape[0]):
        for y in range(b.shape[1]):
            minValue = min(b[x, y], g[x, y], r[x, y])
            min_b_g_r[x, y] = minValue

    tmp = (3/sum)*min_b_g_r
    s_tc = ones - tmp

    # 转到HSV空间
    hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
    h1, s1, v1 = cv2.split(hsv)

    # 转到HLS空间
    hls = cv2.cvtColor(img, cv2.COLOR_BGR2HLS)  # 这就是ihs颜色空间：
    h, l, s = cv2.split(hls)

    # # 文献中的阈值
    # μ_Clouds = 12.7
    # σ_Clouds = 3.7
    #
    # μ_sky = 45.3
    # σ_sky = 4.4

    # # np.where用法：https://www.zhihu.com/question/62844162
    # clouds_loc = np.where(s < μ_Clouds + 3 * σ_Clouds)  # 厚云
    # sky_loc = np.where(s > μ_sky - 3 * σ_sky)  # 天空
    # unknow_loc = np.where((μ_Clouds + 3 * σ_Clouds <= s) & (s <= μ_sky - 3 * σ_sky))  # 薄云

    # 自己通过样本测得的阈值
    μ_Clouds = 170.9
    σ_Clouds = 2.7

    μ_sky = 111.25
    σ_sky = 3.13

    # np.where用法：https://www.zhihu.com/question/62844162
    clouds_loc = np.where(s > μ_Clouds - 3 * σ_Clouds)  # 厚云
    sky_loc = np.where(s < μ_sky + 3 * σ_sky)  # 天空
    unknow_loc = np.where((μ_Clouds + 3 * σ_Clouds <= s) & (s <= μ_sky - 3 * σ_sky))  # 薄云

    s[sky_loc] = 1
    s[clouds_loc] = 2
    s[unknow_loc] = 3

    tmp = image_path.split('.')
    out = tmp[0] + "_saturation." + tmp[1]
    cv2.imwrite(out, s)

    # cv2.imshow('image', cloud_BINARY)
    cv2.waitKey(0)
    cv2.destroyAllWindows()


def calculate_mean_var(image_path):
    img = cv2.imread(image_path)

    hls = cv2.cvtColor(img, cv2.COLOR_BGR2HLS)  # 这就是ihs颜色空间：
    h, l, s = cv2.split(hls)

    mean = np.mean(s)
    var = np.var(s)

    print(image_path)
    print("\n")
    print("均值为：%f\n" % mean)
    print("方差为：%f\n" % var)


if __name__ == '__main__':
    train_image_path1 = r'D:\PycharmProjects\cloudage\train\sky.jpg'
    train_image_path2 = r'D:\PycharmProjects\cloudage\train\cloude.jpg'
    train_image_path3 = r'D:\PycharmProjects\cloudage\train\light cloud.jpg'
    train_image_path4 = r'D:\PycharmProjects\cloudage\train\s61.jpg'
    train_image_path5 = r'D:\PycharmProjects\cloudage\train\s96.jpg'
    # calculate_mean_var(train_image_path1)
    # calculate_mean_var(train_image_path2)
    # calculate_mean_var(train_image_path3)

    # 证明python-opencv的饱和度值域范围未0-255
    # calculate_mean_var(train_image_path4)
    # calculate_mean_var(train_image_path5)

    image_path = r'D:\PycharmProjects\cloudage\photo\photo12.jpg'
    # image_path = r'D:\PycharmProjects\cloudage\photo\cloud1_perfect.jpg'
    # ratio_method1(image_path)
    ratio_method2(image_path)
    # ratio_method1_subtract_ratio_method2(image_path)

    # saturability_method(image_path)