python+opencv图片模糊/过爆/过暗检测一条龙

目前bd到的基本都是c++版本，自己改了一个python的，阈值可以自己微调一下，有问题欢迎指出。
说明有空再写（）
check.py主文件

from PIL import Image
import os
from os import listdir
from tools import calLaplacianVar,overExposeDetect

def check_image(img_path, fixedSize, stdTimes = 1):
    #正常值区间
    # distArea = []
    # distArea.append(calAvg(img) - calStd(img)*stdTimes)
    # distArea.append(calAvg(img) + calStd(img)*stdTimes)
    status = "normal"
    #当图片指标不在正常范围内,被判断为异常
    value = calLaplacianVar(img_path, fixedSize)
    if (value < 2200):
        isOutlier = True
        status = "vague"
        print(str(img_path) + "该图指标值:" + str(value) + " status:" + status)
        print(" 正常区间:[2200 , +)")
        return isOutlier
    else:
        isOutlier = False
    # print("该图指标值:" + str(value) + " status:" + status)
    isOutlier = overExposeDetect(img_path, fixedSize)
    return isOutlier

def main():
    test_folder = r'E:\newproject\test'
    #generate_folder = r'E:\newproject\generate'
    files = listdir(test_folder)
    for f in files:
        # print("图片" + f +":")
        check_image(os.path.join(test_folder, f), (448, 448))
if __name__ == '__main__':
    main()

工具文件tools.py

import cv2
import os
import numpy as np
import math

'''
@illustrate：用拉普拉斯计算图片模糊程度
@param img 图片路径， size(tuple)resize大小
@return 模糊程度
'''
def calLaplacianVar(img, size):
    grayImg = cv2.imread(img, 0)
    grayImg = cv2.resize(grayImg, size)
    sobelImg = cv2.Laplacian(grayImg, cv2.CV_64FC1)
    #标准差
    mu, sigma = cv2.meanStdDev(sobelImg) 
    sigmaValue = sigma[0][0]
    #方差
    variance = pow(sigmaValue, 2)
    return variance

"""
@illustrate: 曝光检测
@param
"""
def overExposeDetect(img_path,size):
    img = cv2.imread(img_path, 1)
    img = cv2.resize(img, size)
    thre = 0.175
    status = "normal"
    flag = False
    if img.shape[2] != 1:
        hsvSpaceImage = cv2.cvtColor(img, cv2.COLOR_RGB2HSV) # hsv转换
    else:
        hsvSpaceImage = img.clone()
    hsvImageVChannels = hsvSpaceImage[:, :, 2]
    step = 8   #以8*8小窗口遍历V通道图像
    imageOverExposeBlockNum = 0
    imageBlocks = 0
    imageDarkBlockNum = 0
    #遍历
    i = 0
    while i < hsvImageVChannels.shape[0]:
        j = 0
        while j < hsvImageVChannels.shape[1]:
            imageBlock = hsvImageVChannels[i:i+step, j:j+step]
            mea = np.mean(imageBlock)# 求小矩形的均值
            if mea > 233.0:
                imageOverExposeBlockNum += 1
            elif mea < 53.0:
                imageDarkBlockNum += 1
            imageBlocks += 1
            j += step
        i += step
    if imageDarkBlockNum/imageBlocks > thre:
        status = "dark"
        flag = True
    if imageOverExposeBlockNum/imageBlocks > thre:
        status = "overexposure"
        flag = True
    if flag == True:
        print(str(img_path) + "该图曝光度:" + str(imageOverExposeBlockNum/imageBlocks * 100) + " status:" + status)
        print(" 正常区间:(0," + str(thre*100) + "]")
    return flag