强大的Python图像处理
import os
import sys
import subprocess
from pytesser_pro.pytesser_pro import *
import Image, ImageEnhance, ImageFilter
from pylab import *
# 二值化并转格式
def binary(image_name, binary_image_name):
# 白底黑字
args="convert -monochrome "+image_name+" "+binary_image_name
# print args
proc=subprocess.Popen(args, shell=True)
proc.wait()
im=Image.open(binary_image_name)
w, h=im.size
data=list(im.getdata())
if (data[0], data[w-1], data[(h-1)*w], data[h*w-1])==(0, 0, 0, 0): # 0-黑色,255-白色
# 若非白底黑字则灰度反转
args1="convert -negate "+binary_image_name+" "+binary_image_name
proc1=subprocess.Popen(args1, shell=True)
proc1.wait()
# 计算范围内点的个数
def numpoint(im):
w, h=im.size
# print w, h
data=list(im.getdata())
mumpoint=0
for x in range(w):
for y in range(h):
if data[y*w+x]==0: # 0-黑色,255-白色
mumpoint +=1
return mumpoint
# 投影法去干扰线
def pointmidu(binary_image_name, midu_image_name):
im=Image.open(binary_image_name)
w, h=im.size
# print w, h
len=5
for x in range(0, w, len):
box=(x, 0, x+len, h)
im_box=im.crop(box)
num=numpoint(im_box)
# print num
if num
for i in range(x, x+len):
for j in range(h):
im.putpixel((i, j), 255) # 0-黑色,255-白色
data=list(im.getdata())
data_column=[]
for x in range(w):
temp=0
for y in range(h):
if data[y*w+x]==0: # 0-黑色,255-白色
temp +=1
data_column.append(temp)
# print data_column
start=0
for i in range(0, w, 1):
if data_column[i] !=0:
break
else:
start +=1
# print start
end=w-1
for j in range(w-1, -1, -1):
if data_column[j] !=0:
break
else:
end +=-1
# print end
box_new=(start, 0, end+1, h)
im_box_new=im.crop(box_new)
im_box_new.save(midu_image_name)
# 图像增强
def filter_enhance(midu_image_name, midu_image_name_pro1):
im=Image.open(midu_image_name)
# 去噪
im=im.filter(ImageFilter.MedianFilter())
# 亮度加强
enhancer=ImageEnhance.Contrast(im)
im=enhancer.enhance(2)
im=im.convert('1')
# im()
im.save(midu_image_name_pro1)
# 字符分割
def seg(midu_image_name_pro1, midu_image_name_pro2, num):
im=Image.open(midu_image_name_pro1)
w, h=im.size
# print w, h, w/num
len=2
for i in range(num-1):
start=(i+1)*w/num
end=start+len
for m in range(start, end+1):
for n in range(h):
im.putpixel((m, n), 255) # 0-黑色,255-白色
im.save(midu_image_name_pro2)
def get_aim1_point(im):
aim=[]
w, h=im.size
# print w, h
data=list(im.getdata())
for x in range(0, w, 1):
for y in range(0, h, 1):
if data[y*w+x]==0: # 0-黑色,255-白色
start_point=(x, y)
# print start_point
aim.append(start_point)
break
return aim
def get_aim2_point(im):
aim=[]
w, h=im.size
# print w, h
data=list(im.getdata())
for x in range(0, w, 1):
for y in range(h-1, -1, -1):
if data[y*w+x]==0: # 0-黑色,255-白色
start_point=(x, y)
# print start_point
aim.append(start_point)
break
return aim
if __name__=='__main__':
if len(sys.argv)==1:
image_name="./pic/get_random.jpg" # QQ账号买号验证码图片名称
digits=False
# image_name="./pic/get_price_img.png" # 价格图片名称
# digits=True
elif len(sys.argv)==2:
if sys.argv[1].find("get_random") !=-1:
image_name=sys.argv[1]
digits=False
elif sys.argv[1].find("get_price_img") !=-1:
image_name=sys.argv[1]
digits=True
else:
print "Please Input the Correct Image Name!"
sys.exit(0)
else:
print "Too Many Arguments!"
sys.exit(0)
# 二值化并转格式
binary_image_name=os.path.splitext(image_name)[0]+"_binary.png"
binary(image_name, binary_image_name)
im=Image.open(binary_image_name)
print im.format, im.size, im.mode
if digits:
text=image_file_to_string(binary_image_name, bool_digits=digits)
print text.replace("
", "")
else:
# 投影法去干扰线
fpathandname , fext=os.path.splitext(binary_image_name)
midu_image_name=fpathandname+"_midu"+fext
pointmidu(binary_image_name, midu_image_name)
fpathandname , fext=os.path.splitext(midu_image_name)
# 去干扰线
# im=Image.open(midu_image_name)
# w, h=im.size
# data=list(im.getdata())
# aim1=get_aim1_point(im)
# for x, y in aim1:
# curr=data[y*w+x]
# prev=data[(y-1)*w+x]
# next=data[(y+1)*w+x]
#
# if prev==0 and next==0: # 0-黑色,255-白色
# continue
# if prev==0:
# im.putpixel((x, y), 255)
# im.putpixel((x, y-1), 255)
# elif next==0:
# im.putpixel((x, y), 255)
# im.putpixel((x, y+1), 255)
# else:
# im.putpixel((x, y), 255)
# data=list(im.getdata())
# aim2=get_aim2_point(im)
# for x, y in aim2:
# curr=data[y*w+x]
# prev=data[(y-1)*w+x]
# next=data[(y+1)*w+x]
#
# if prev==0 and next==0: # 0-黑色,255-白色
# continue
# if prev==0:
# im.putpixel((x, y), 255)
# im.putpixel((x, y-1), 255)
# elif next==0:
# im.putpixel((x, y), 255)
# im.putpixel((x, y+1), 255)
# else:
# im.putpixel((x, y), 255)
# midu_image_name_new=fpathandname+"_new"+fext
# im.save(midu_image_name_new)
# 图像增强
midu_image_name_pro1=fpathandname+"_pro1"+fext
filter_enhance(midu_image_name, midu_image_name_pro1)
# 字符分割
# num=4
# midu_image_name_pro2=fpathandname+"_pro2"+fext
# seg(midu_image_name_pro1, midu_image_name_pro2, num)
# im=Image.open(midu_image_name)
# text=image_to_string(im)
# print text.replace("
", "")
text=image_file_to_string(midu_image_name_pro1, bool_digits=digits)
print text.replace("
", "")
上一篇: 降低Java垃圾回收开销的一些建议