opencv-身份证文字提取

# -*- coding: UTF-8 -*-

import cv2
import numpy as np

def get_WH(w, h):
    return int(w*0.63),int(h*0.8)

def get_image(image):
    h,w = image.shape
    w2, h2 = get_WH(w, h)
    for x in range(w2,w):
        for y in range(1,h2):
            image[y-1,x-1] = 0
    return image

def get_ranges(array, min_s = 0, color_size = 255, min_col = None):
    rs = []
    ranges = []
    start = None
    end = None
    for i,s in enumerate(array):
        s = s / color_size
        if(s > min_s) :
            if(start is None) :
                start = i
            elif (start is not None) :
                end = i
        if(s <= min_s and end is not None and start is not None) :
            if(end is not None and start is not None 
               and min_col is not None and (end - start) < min_col):
                continue
            rs.append((start, end - 1))
            ranges.append(end - start)
            start = None
            end = None
        if(i == len(array) - 1 and start is not None and end is not None):
            rs.append((start, len(array) - 1))
            
    return rs,ranges

def get_show_image(img_range, image):
    color = (255, 0, 0)
    for i,r in enumerate(img_range):
        x,y,w,h = r
        pt1 = (x, y)
        pt2 = (x + w, y + h)
        cv2.rectangle(image, pt1, pt2, color)
    return image

def get_cols(image, l, cols, col_ranges):
    (min_y,max_y) = l
    cols2 = []
    if(col_ranges is None or len(col_ranges) < 1):
        return cols2
    a = sum(col_ranges) / len(col_ranges)
    deviation = int(a * 0.5)
    for i,r in enumerate(col_ranges):
        (min_x,max_x) = cols[i]
        if((r - a) > deviation):
            line_img = image[min_y:max_y, min_x:max_x]
            line_img_sum = np.sum(line_img, axis=0)
            cols_new,col_ranges_new = get_ranges(line_img_sum, min_s=2, min_col= a - deviation)
            for j,c in enumerate(cols_new):
                cols2.append((min_x+cols_new[j][0], min_x+cols_new[j][1]))
        else:
            cols2.append(cols[i])
    return cols2

def get_img_range(image, percent=None):
    line_sum = np.sum(image, axis=1)
    # print(line_sum)
    lines,ranges = get_ranges(line_sum)
    img_range = []
    for i,l in enumerate(lines):
        (min_y,max_y) = l
        line_img = image[min_y:max_y, :]
        line_img_sum = np.sum(line_img, axis=0)
        cols,col_ranges = get_ranges(line_img_sum)
        cols = get_cols(image, l, cols, col_ranges)
        cols = get_cols_by_min(cols, col_ranges, percent)
        for j,c in enumerate(cols):
            (min_x,max_x) = c
            img_range.append((min_x, min_y, (max_x - min_x), (max_y - min_y)))
    return img_range

def get_ranges_by_col(img_range, interval_s = 0):
    if(interval_s == 0):
        return range
    old_x = None
    old_y = None
    range2 = []
    for (i,r) in enumerate(img_range):
        x,y,w,h = r
        if(i == len(img_range) - 1):
            if(old_x is not None and old_y is not None):
                range2.append((old_x, old_y, x - old_x + w, h))
                old_x = None
                old_y = None
            else:
                range2.append((x,y,w,h))
        else:
            if(old_x is not None and old_y is not None):
                range2.append((old_x, old_y, x - old_x + w, h))
                old_x = None
                old_y = None
                continue
            x2,y2,w2,h2 = img_range[i+1]
            interval = x2 - x
            if(interval > 0 and interval < interval_s):
                old_x = x
                old_y = y
            else:
                range2.append((x,y,w,h))
                old_x = None
                old_y = None
    return range2
    
def get_cols_by_min(cols, col_ranges, percent=None):
    if(percent is None):
        return cols
    cols2 = []
    if(col_ranges is None or len(col_ranges) < 1):
        return cols2
    a = sum(col_ranges) / len(col_ranges)
    deviation = int(a * percent)
    for i,r in enumerate(col_ranges):
        (min_x,max_x) = cols[i]
        if((a - r) > deviation):
            if(i < len(col_ranges) - 1):
                cols2.append((min_x, cols[i+1][1]))
                i+=1
        else:
            cols2.append(cols[i])
    return cols2

path = "demo5_4.png"
image_color = cv2.imread(path)
new_shape = (image_color.shape[1] * 2, image_color.shape[0] * 2)
image_color = cv2.resize(image_color, new_shape)
lower = np.array((0, 0, 0), dtype = "uint8")
upper = np.array((100, 100, 100), dtype = "uint8")
image = cv2.inRange(image_color, lower, upper)

image = get_image(image)

img_range = get_img_range(image)
img_range = get_ranges_by_col(img_range, interval_s=15)
image = get_show_image(img_range, image_color)
     
cv2.imshow('image', image)
cv2.waitKey(0)

提取效果：