帮忙写个代码，可以识别框选字体部分（矩形框），并分割成单独图片

Question

 

Answer 1

python实现，需要安装opencv、numpy模块：切割准确率95%以上， 效果图：

 

#juzicode.com/vx:桔子code
import os,sys,time,cv2
import numpy as np

dbg_is_show = False
def show_img(win_name,img,wait_time=0,img_ratio=0.15,is_show=True):
    if is_show is not True:
        return 
    rows = img.shape[0]
    cols = img.shape[1]
    cv2.namedWindow(win_name, cv2.WINDOW_NORMAL )#cv2.WINDOW_AUTOSIZE)
    cv2.resizeWindow(win_name,(int(cols*img_ratio),int(rows*img_ratio)))
    cv2.imshow(win_name,img)
    cv2.waitKey(wait_time)
    
if not os.path.exists('out'):
    os.mkdir('out')
    
print('juzicode.com/vx:桔子code')
print(cv2.__version__)
img_src = cv2.imread('src.jpg')
print(img_src.shape)

show_img('img_src',img_src,is_show=dbg_is_show)
 
#获取灰度图
img_b, img_g, img_r = cv2.split(img_src) 
show_img('img_r',img_r,is_show=dbg_is_show)

img_gray = cv2.bitwise_not(img_r)
img_gray= cv2.medianBlur(img_gray,5)

#二值化
thresh_bin,img_bin= cv2.threshold(img_gray,127,255,cv2.THRESH_BINARY)
show_img('img_bin',img_bin,is_show=dbg_is_show)

kernel = cv2.getStructuringElement(cv2.MORPH_RECT,(3, 3))
img_eroded = cv2.erode(img_bin,kernel)      
show_img('img_eroded',img_eroded)   

kernel = cv2.getStructuringElement(cv2.MORPH_RECT,(29, 29))
img_dilated = cv2.dilate(img_eroded,kernel)       
show_img('img_dilated',img_dilated)

res = cv2.findContours(img_dilated,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)  
contours=res[1] 
print ('len(contours):',len(contours))

for i in range(0,len(contours)): 
    x, y, w, h = cv2.boundingRect(contours[i])
    print(i,len(contours[i]))
    if len(contours[i])<80:continue
    cv2.rectangle(img_src, (x,y), (x+w,y+h), (255,0,0), 10)
    newimage=img_src[y:y+h,x:x+w]
    cv2.imwrite( 'out\\'+str(i)+".jpg",newimage)
    
show_img("img_dilated_with_contour", img_src)