Python代码库OpenCV之13 古书一页切成两页
前置项目
首先将pdf切割成图片，
- Python代码库之PDF转换为图片poppler
将图片切割为2页 - Python代码库OpenCV之13 古书一页切成两页
测试图片
1.png
代码
import cv2
import numpy as np
import cv2
from matplotlib import pyplot as plt
import os
import pandas as pd
import shutil
img_filename='./image/1.png'
out_path='./chunqiu_out_split'
filename=0
# 二值化
def thresh_binary(img):
blur = cv2.GaussianBlur(img, (9, 9), 0)
# OTSU's binaryzation
ret3, th3 = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)
kernel = np.ones((2, 2), np.uint8)
opening = cv2.morphologyEx(th3, cv2.MORPH_OPEN, kernel)
return opening
# 统计每一行的黑点数
def hist_col(img):
list=[]
row, col = img.shape
for i in range(col):
list.append((img[:, i] < 200).sum())
return list
def cut_col(img, hist_list):
minlist = []
images = []
row, col = img.shape
np_list = np.array(hist_list)
avg = col/64
i = 0
# print np_list
while i < col-1:
if i >= col-10:
if np_list[i] < 40 and np_list[i] <= np_list[i+1: col].min():
minlist.append(i)
break
if i == col-1:
minlist.append(i)
break
else:
i=int(i)
if np_list[i] < 40 and np_list[i] < np_list[i+1: i+10].min():
minlist.append(i)
i += avg
i += 1
print (minlist)
for j in range(len(minlist)-1):
print (j)
images.append(img[0:row, minlist[j]:minlist[j+1]])
return images
# 直方图绘制
def hist_show(one_list):
data=one_list
# 将列表转换成pandas的Series格式(因为里边有自带的函数可以统计元素出现的个数， for统计列表中元素出现的次数，效率过慢，不适用于大数据量)
data_pd = pd.Series(data)
l_unique_data = list(data_pd.value_counts().index) # 该数据中的唯一值
l_num = list(data_pd.value_counts()) # 唯一值出现的次数
plt.bar(l_unique_data, l_num, width=0.1)
plt.show()
# sum the black pixel numbers in each rows
def hist_row(img):
list=[]
row, col = img.shape
for i in range(row):
list.append((img[i, :] < 200).sum())
return cut_row(img, list)
# find each segmentation rows
def cut_row(img, row_list):
minlist = []
single_images_with_rect = []
row, col = img.shape
np_list = np.array(row_list)
avg = row/32
i = 0
while i <= row-1:
i=int(i)
if i >= row-10 and np_list[i] == 0:
minlist.append(i)
break
elif np_list[i] == 0 and (np_list[i+1: i+10] < 200).sum() >= 5:
minlist.append(i)
i += avg
i += 1
print (minlist)
for j in range(len(minlist)-1):
single_img = img[minlist[j]:minlist[j+1], 0:col]
single_img_with_rect = single_cut(single_img)
if single_img_with_rect is not None:
single_images_with_rect.append(single_img_with_rect)
return single_images_with_rect
# find the single word's contours and take off the redundant margin
def single_cut(img):
global filename
blur = cv2.GaussianBlur(img, (9, 9), 0)
ret3, th3 = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)
contours, hierarchy = cv2.findContours(th3, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
up, left = img.shape
down, right = 0, 0
for i in range(len(contours)):
cnt = contours[i]
x, y, w, h = cv2.boundingRect(cnt)
if w < 6 and h < 6:
continue
if x < up:
up = x
if y < left:
left = y
if x+w > down:
down = x+w
if y+h > right:
right = y+h
if down-up >= 40 and right-left >= 40:
word = img[left:right, up:down]
cv2.imwrite(out_path+'/'+str(filename)+'.png', word)
cv2.rectangle(img,(up, left), (down, right), (0, 255, 0), 2)
filename += 1
return img
else:
return None
# 准备工作，文件
##
## 创建文件夹
if os.path.exists(out_path):
shutil.rmtree(out_path, ignore_errors=True)
os.mkdir(out_path)
#第一步，灰度
img = cv2.imread(img_filename,cv2.IMREAD_GRAYSCALE)
#第二步 二值化（膨胀blur，腐蚀）
opening=thresh_binary(img)
#反转
r,c=0,0
for rows in opening:
c=0
for citem in rows:
#print('r',r,'c',c)
if citem == 255:opening[r,c]=0
else: opening[r,c]=255
c=c+1
r=r+1
#第三步，直方图
hist_list = hist_col(opening)
#第四步，切割图片
images = cut_col(opening, hist_list)
# create two plt
fig, axes = plt.subplots(1, len(images), sharex=True, sharey=True)
fig2, axes2 = plt.subplots(len(images), 12, sharex=True, sharey=True)
#第五步，
# split the pieces into single words by rows
for i in range(len(images)):
axes[i].imshow(images[i], 'gray')
single_images_with_rect = hist_row(images[i])
for j in range(len(single_images_with_rect)):
axes2[i, j].imshow(single_images_with_rect[j], 'gray')
fig.savefig(out_path+'/cut_col.png')
fig2.savefig(out_path+'/single.png')
plt.clf()
#cv2.imshow('gray',img)
#cv2.waitKey()
#cv2.imshow('opening1',opening)
#cv2.waitKey(0)
更多精彩代码请关注我的专栏
- selenium & python 源码大全
- reportlab教程和源码大全
- python源码大全