图片验证码预处理

Posted 2021-03-26 chenxiyuxiao

from copy import deepcopy

from PIL import Image
import numpy as np
from collections import Counter
import hashlib


def fileMd5(filePath):
    with open(filePath, ‘rb‘) as f:
        a = hashlib.md5()
        a.update(f.read())

        return a.hexdigest()


# 图片只有背景色与文字色，两个颜色的处理  二值化处理,图片颜色越单调，效果越好
# 可用来处理大部分彩色图片，非干扰线
def parseImg_1(filePath):
    img = Image.open(filePath)
    lim = img.convert(‘L‘)
    data = np.asarray(lim)
    onea = data.reshape(np.multiply(*data.shape))
    d_dict = dict(Counter(onea))
    keys = list(d_dict.keys())
    sort_ = lambda x: d_dict.get(x)
    keys.sort(key=sort_, reverse=True)
    ress = []
    for z in keys:
        if not ress:
            ress.append(int(z))
        elif abs(z - ress[0]) > 50:
            ress.append(int(z))
            break
    threshold = np.mean(ress)
    dd = np.where(data > threshold, 255, 0)
    img2 = Image.fromarray(np.uint8(dd))
    img2.save(‘parsed.jpg‘)


# 背景色转白色
def parseImg_2(filePath, transNum=35):
    """

    :param filePath: 图片路径
    :param transNum: 根据背景色的多少，若只有一种 ，则写1 即可
    :return:
    """
    img = Image.open(filePath)
    img = img.convert(‘L‘)
    data = np.asarray(img)
    ddd = data.reshape(np.multiply(*data.shape))
    k = dict(Counter(ddd))
    keys = list(k.keys())
    sort_ = lambda x: k.get(x)
    keys.sort(key=sort_, reverse=True)
    # zh = keys[0]
    for ind in range(transNum):
        data = np.where(data == keys[ind], 255, data)
    # data = np.where(data>,data,255)
    img2 = Image.fromarray(np.uint8(data))
    img2.save(‘parsed.jpg‘)


# 干扰线降噪 适合处理较细的线，如果线太粗则效果很糟糕 , 处理一次往往得不到最好的效果
# 也可用于去除噪点 ，太粗的点无法去除
def parseImg_3(filePath):
    img = Image.open(filePath)
    img = img.convert(‘L‘)
    img = np.asarray(img)
    # img = np.ndarray(img)
    ano = np.ndarray(shape=img.shape)
    h, w = img.shape[:2]
    # ！！！opencv矩阵点是反的
    # img[1,2] 1:图片的高度，2：图片的宽度
    for y in range(1, w - 1):
        for x in range(1, h - 1):
            count = 0
            if img[x, y - 1] > 150:
                count = count + 1
            if img[x, y + 1] > 150:
                count = count + 1
            if img[x - 1, y] > 150:
                count = count + 1
            if img[x + 1, y] > 150:
                count = count + 1
            if count > 2:
                # img[x, y] = 255
                ano[x, y] = 255
            else:
                ano[x, y] = img[x, y]
    ano = np.where(ano > 150, 255, 0)
    img2 = Image.fromarray(np.uint8(ano))
    img2.save(filePath)


# 干扰线降噪， 多次调用 第三个处理方法，直到处理干净为止
def parseImg_3_wanshanbanben(filepath):
    first = fileMd5(filepath)
    while True:
        parseImg_3(filepath)
        second = fileMd5(filepath)
        if first == second:
            break
        else:
            first = second

# 空心转实心算法
def parseImg_4(filePath):
    img = Image.open(filePath)
    img = img.convert(‘L‘)
    img = np.asarray(img)
    img = np.where(img > 200, 255, 0)
    duizhao = np.zeros(shape=img.shape)
    w,h = img.shape
    # print(w,h)
    isWhite = lambda x: img[x[0],x[1]]==255
    startZuobiao = [[0, 0],[0,h],[w,0],[w,h] ]
    hasPanduan = set()
    hasPed = lambda x: ‘*‘.join([str(item) for item in x]) in hasPanduan
    hasPed_add = lambda x: hasPanduan.add(‘*‘.join([str(item) for item in x]))
    totalLen = deepcopy(len(hasPanduan))
    # print(totalLen)
    while True:
        for zuo in startZuobiao:
            x,y = zuo
            for i,j in [
                [x,y+1],[x,y-1],[x+1,y],[x-1,y] ,[x-1,y-1],[x-1,y+1],[x+1,y-1],[x+1,y+1]
            ]:
                if 0<=i<=w-1 and 0<=j<=h-1:
                    if not hasPed([i,j]) and    isWhite([i,j]):
                        duizhao[i,j] = 255
                        startZuobiao.append([i,j])
                    hasPed_add([i,j])
        if len(hasPanduan) == totalLen:
            # print(totalLen)
            break
        else:
            totalLen = deepcopy(len(hasPanduan))






    img2 = Image.fromarray(np.uint8(duizhao))
    img2.save(‘parsed.jpg‘)


if __name__ == ‘__main__‘:
    # parseImg_1(‘pin.png‘)
    # parseImg_1(‘q.jpg‘)
    # # parseImg_2(‘ws.jpg‘)
    # # parseImg_3(‘parsed.jpg‘)
    # parseImg_3_wanshanbanben(‘parsed.jpg‘)
    parseImg_4(‘pin.png‘)
    # parseImg_1(‘pin2.png‘)
    # parseImg_3_wanshanbanben(‘parsed.jpg‘)