opencv_文档ocr扫描

总体步骤：

1. 图片预处理,
2. 透视变换, 视角拉正.
3. tesseract进行识别.

import numpy as np
import cv2 

图片预处理

# 读取图片
image = cv2.imread('./images/page.jpg')

# 计算比例. 限定高度500
ratio = image.shape[0] / 500.0

# 保存原始大小的图片
orig = image.copy()

# 修改resize，使其按比例缩放
# 封装resize功能 -- width，height只传入其一
def resize(image, width=None, height=None, inter=cv2.INTER_AREA):
    # 保存修改的宽高
    dim = None
    
    # 获取图片的宽高
    (h, w) = image.shape[:2]
    
    # 宽高均未指定，返回原图像
    if width is None and height is None:
        return image
    
    # 指定了height
    if width is None:
        r = height / float(h) # 计算比例
        dim = (int(w * r), height)
        
    # 指定了width
    else:
        r = width / float(w)
        dim = (width, int(h * r))
    
    # 按比例缩放，并返回
    resized = cv2.resize(image, dim, interpolation=inter)
    return resized

# 对图片进行resize，指定height为500
image = resize(orig, height=500)

image.shape

(500, 375, 3)

# 显示窗口
def cv_show(name, image):
    cv2.imshow(name, image)
    cv2.waitKey(0)
    cv2.destroyAllWindows()
    cv2.waitKey(1)

# 图片预处理
# 灰度化处理
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

# 高斯平滑 -- 去除噪声
gray = cv2.GaussianBlur(gray, (5, 5), 0)

# 边缘检测
edged = cv2.Canny(gray, 75, 200)

cv_show('edged', edged)

# 轮廓检测并返回
cnts = cv2.findContours(edged.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)[0]

# 轮廓，按照面积排序，降序排序
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)
cnts

# 画出轮廓
image_contours = cv2.drawContours(image.copy(), cnts, -1, (0, 0, 255), 2)
cv_show('image_contours', image_contours)

# 遍历轮廓找出最大的轮廓. 
for c in cnts:
    # 计算轮廓周长
    perimeter = cv2.arcLength(c, True)
    
    # 多边形逼近.
    approx = cv2.approxPolyDP(c, 0.02 * perimeter, True)
    
    if len(approx) == 4:
        screen_cnt = approx # screen_cnt为矩形轮廓
        break

image_contours = cv2.drawContours(image.copy(), [screen_cnt], -1, (0, 0, 255), 2)
cv_show('image_contours', image_contours)

透视变换

# 对角点坐标进行排序，顺时针排序
def order_points(pts):
    # 创建全是0的矩阵, 来接收等下找出来的4个角的坐标.
    rect = np.zeros((4, 2), dtype='float32')
    
    # s存储各角点的x, y坐标之和
    s = pts.sum(axis=1)
    
    # 左上的坐标一定是x,y加起来最小的坐标. 右下的坐标一定是x,y加起来最大的坐标.
    rect[0] = pts[np.argmin(s)]
    rect[2] = pts[np.argmax(s)]
    
    # diff存储各角点的x, y坐标之差
    diff = np.diff(pts, axis=1)
    
    # 右上角的x,y相减的差值一定是最小的. 
    # 左下角的x,y相减的差值, 一定是最大.
    rect[1] = pts[np.argmin(diff)]
    rect[3] = pts[np.argmax(diff)]
    
    return rect

# 把透视变换功能封装成一个函数
def four_point_transform(image, pts):
    # 对输入的4个坐标排序
    rect = order_points(pts)
    
    # 返回排序后的结果
    (tl, tr, br, bl) = rect
    
    # 空间中两点的距离
    widthA = np.sqrt((br[0] - bl[0]) ** 2 + (br[1] - bl[1]) ** 2)
    widthB = np.sqrt((tr[0] - tl[0]) ** 2 + (tr[1] - tl[1]) ** 2)
    max_width = max(int(widthA), int(widthB))
    
    heightA = np.sqrt((tr[0] - br[0]) ** 2 + (tr[1] - br[1]) ** 2)
    heightB = np.sqrt((tl[0] - bl[0]) ** 2 + (tl[1] - bl[1]) ** 2)
    max_height = max(int(heightA), int(heightB))
    
    # 构造变换之后的对应坐标位置.
    dst = np.array([
        [0, 0],
        [max_width - 1, 0],
        [max_width - 1, max_height - 1],
        [0, max_height - 1]], dtype='float32')
    
    # 计算变换矩阵
    M = cv2.getPerspectiveTransform(rect, dst)
    
    # 透视变换，传入变换之后的图片大小
    warped = cv2.warpPerspective(image, M, (max_width, max_height))
    return warped

# 对原始大小的图片，进行透视变换
warped = four_point_transform(orig, screen_cnt.reshape(4, 2) * ratio)

warped.shape

(2624, 1962, 3)

cv_show('warped', warped)

# 灰度化处理
warped = cv2.cvtColor(warped, cv2.COLOR_BGR2GRAY)

# 二值化处理
ref = cv2.threshold(warped, 100, 255, cv2.THRESH_BINARY)[1]

cv_show('ref', ref)

# 把处理好的图片写入图片文件.
cv2.imwrite('./scan.jpg', ref)

True

tesseract进行识别

# 导包
import pytesseract
from PIL import Image

# pytesseract要求的image：不是opencv读进来的image, 而是pillow这个包, 即PIL读进来的image
text = pytesseract.image_to_string(Image.open('./scan.jpg'))
print(text)