使用Python+OpenCV2进行图片中的文字分割（支持竖版）

news2025/1/21 20:22:04

扣字和分割

把图片中的文字，识别出来，并将每个字的图片抠出来；

import cv2
import numpy as np

HIOG = 50
VIOG = 3
Position = []
 
'''水平投影'''
def getHProjection(image):
    hProjection = np.zeros(image.shape,np.uint8)
    # 获取图像大小
    (h,w)=image.shape
    # 统计像素个数
    h_ = [0]*h
    for y in range(h):
        for x in range(w):
            if image[y,x] == 255:
                h_[y]+=1
    #绘制水平投影图像
    for y in range(h):
        for x in range(h_[y]):
            hProjection[y,x] = 255
    # cv2.imshow('hProjection2',cv2.resize(hProjection, None, fx=0.3, fy=0.5, interpolation=cv2.INTER_AREA))
    # cv2.waitKey(0)
    return h_
 
def getVProjection(image):
    vProjection = np.zeros(image.shape,np.uint8);
    (h,w) = image.shape
    w_ = [0]*w
    for x in range(w):
        for y in range(h):
            if image[y,x] == 255:
                w_[x]+=1
    for x in range(w):
        for y in range(h-w_[x],h):
            vProjection[y,x] = 255
    # cv2.imshow('vProjection',cv2.resize(vProjection, None, fx=1, fy=0.1, interpolation=cv2.INTER_AREA))
    # cv2.waitKey(0)
    return w_


def scan(vProjection, iog, pos = 0):
    start = 0
    V_start = []
    V_end = []

    for i in range(len(vProjection)):
        if vProjection[i] > iog and start == 0:
            V_start.append(i)
            start = 1
        if vProjection[i] <= iog and start == 1:
            if i - V_start[-1] < pos:
                continue
            V_end.append(i)
            start = 0
    return V_start, V_end


def checkSingle(image):
    h = getHProjection(image)
    start = 0
    end = 0

    for i in range(h):
        pass

#分割
def CropImage(image,dest,boxMin,boxMax):
    a=boxMin[1]
    b=boxMax[1]
    c=boxMin[0]
    d=boxMax[0]
    cropImg = image[a:b,c:d]
    cv2.imwrite(dest,cropImg)
    
#开始识别
def DOIT(rawPic):
        # 读入原始图像
    origineImage = cv2.imread(rawPic)
    # 图像灰度化   
    #image = cv2.imread('test.jpg',0)
    image = cv2.cvtColor(origineImage,cv2.COLOR_BGR2GRAY)
    
    # cv2.imshow('gray',image)
    # 将图片二值化
    retval, img = cv2.threshold(image,127,255,cv2.THRESH_BINARY_INV)
    # kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (5, 5))
    # img = cv2.erode(img, kernel)
    # cv2.imshow('binary',cv2.resize(img, None, fx=0.3, fy=0.3, interpolation=cv2.INTER_AREA))
    #图像高与宽
    (h,w)=img.shape
    #垂直投影
    V = getVProjection(img)
 
    start = 0
    V_start = []
    V_end = []

    # 对垂直投影水平分割
    V_start, V_end = scan(V, HIOG)
    if len(V_start) > len(V_end):
        V_end.append(w-5)

    # 分割行，分割之后再进行列分割并保存分割位置
    for i in range(len(V_end)):
        #获取行图像
        if V_end[i] - V_start[i] < 30:
            continue

        cropImg = img[0:h, V_start[i]:V_end[i]]
        # cv2.imshow('cropImg',cropImg)
        # cv2.waitKey(0)
        #对行图像进行垂直投影
        H = getHProjection(cropImg)
        H_start, H_end = scan(H, VIOG, 40)


        if len(H_start) > len(H_end):
            H_end.append(h-5)

        for pos in range(len(H_start)):
            # 再进行一次列扫描
            DcropImg = cropImg[H_start[pos]:H_end[pos], 0:w]
            d_h, d_w = DcropImg.shape
            # cv2.imshow("dcrop", DcropImg)
            sec_V = getVProjection(DcropImg)
            c1, c2 = scan(sec_V, 0)
            if len(c1) > len(c2):
                c2.append(d_w)

            x = 1
            while x < len(c1):
                if c1[x] - c2[x-1] < 12:
                    c2.pop(x-1)
                    c1.pop(x)
                    x -= 1
                x += 1

            # cv2.waitKey(0)
            if len(c1) == 1:
                Position.append([V_start[i],H_start[pos],V_end[i],H_end[pos]])
            else:
                for x in range(len(c1)):
                    Position.append([V_start[i]+c1[x], H_start[pos],V_start[i]+c2[x], H_end[pos]])

    #根据确定的位置分割字符
    number=0
    for m in range(len(Position)):
        rectMin =  (Position[m][0]-5,Position[m][1]-5)
        rectMax =  (Position[m][2]+5,Position[m][3]+5)
        cv2.rectangle(origineImage,rectMin, rectMax, (0 ,0 ,255), 2)
        number=number+1
        #start-crop
        CropImage(origineImage,'result/' + '%d.jpg' % number,rectMin,rectMax)
    # cv2.imshow('image',cv2.resize(origineImage, None, fx=0.6, fy=0.6, interpolation=cv2.INTER_AREA))
    cv2.imshow('image',origineImage)
    cv2.imwrite('result/' + 'ResultImage.jpg' , origineImage)
    cv2.waitKey(0)

#############################
rawPicPath = r"H:\TEMP\TEXT_PROCCESS\TEST05.jpg"
DOIT(rawPicPath)
#############################

原图片：

分割后文件夹：

重命名

可见此时文件都还是数字为文件名称，那么接下来要利用OCR自动给每个文字图片文件命名

我们使用UMIOCR ， UMI-OCR的安装建议去GITHUB上查，windows上部署还是很方便的；

这里使用本机安装好的UMI-OCR 的 API地址 http://127.0.0.1:1224/api/ocr

先定义API调用方法

  
############################################
import base64
import requests
import json

#API访问使用
def UMI_OCR_OPT(url,img_path): 
    # url = "http://127.0.0.1:1224/api/ocr"
    # img_path= './result/123.jpg'
 
    with open(img_path,'rb') as f:
        image_base64 = base64.b64encode(f.read())
        image_base64 =str(image_base64,'utf-8')
    
    data = {
        "base64": image_base64,
        # 可选参数
        # Paddle引擎模式
        # "options": {
        #     "ocr.language": "models/config_chinese.txt",
        #     "ocr.cls": False,
        #     "ocr.limit_side_len": 960,
        #     "tbpu.parser": "MergeLine",
        # }
        # Rapid引擎模式
        # "options": {
        #     "ocr.language": "简体中文",
        #     "ocr.angle": False,
        #     "ocr.maxSideLen": 1024,
        #     "tbpu.parser": "MergeLine",
        # }
    }
    headers = {"Content-Type": "application/json"}
    data_str = json.dumps(data)
    response = requests.post(url, data=data_str, headers=headers)
    if response.status_code == 200:
        res_dict = json.loads(response.text)
        #检测失败与否
        if(str(res_dict).find('No text found in image')!=-1):
            # print("返回失败内容\n", res_dict)
            return ''
        print("返回值字典\n", res_dict)
        resText = res_dict['data'][0]['text']
        return resText
    return ''

开始批量调用检测

import easyocr

import shutil

error =''

#api地址
uni_orc_url="http://127.0.0.1:1224/api/ocr"
#将所有字体
for i in range(1,285):
    filePath = "./result/"+str(i)+".jpg"
    # result = reader.readtext(filePath, detail = 0)
    result = UMI_OCR_OPT(uni_orc_url,filePath)
    if(len(result)==0 or result == ''):
        error+= filePath +'\n'
        continue
    print(result)
    destPath = "./resultX/"+result[0]+".jpg" 
    print('sour :: '+ filePath)
    print('dest :: '+ destPath)
    shutil.copyfile(filePath, destPath)

print('error:\n'+error)

然后将能够识别出来的所有文字图片都复制并重命名为该字

效果如下：