python对图片的像素识别及处理
滑动验证码位置确定：
图片示例：
from PIL import Image
def recognize_box(img_path, img_width):
"""
滑块验证图片的路径； 图片的大小
:param img_path:
:param img_width:
:return:
"""
im = Image.open(img_path)
im = im.convert('L')
rows, cols = im.size
row_dot_list = []
for i in range(rows):
col_dot_list =[]
for j in range(cols):
if im.getpixel((i,j)) < 50: # 灰度后趋向于黑色，一般不会大于50
col_dot_list.append(j)
col_dot_list.sort()
if len(get_skip_frame(1, col_dot_list)) >= img_width:
row_dot_list.append(i)
row_dot_list.sort()
move_length = get_skip_frame(1, row_dot_list)
if len(move_length) >= img_width:python对图片的像素识别及处理
print(move_length[0])
def get_skip_frame(skip, frame_list):
"""
间隔数， 连续数据列表
:param skip:
:param frame_list:
:return:
"""
tmp_list = []
return_list = []
for i in range(len(frame_list)):
if len(tmp_list) == 0:
tmp_list.append(frame_list[i])
else:
change = abs(frame_list[i] - tmp_list[-1])
if change == skip:
tmp_list.append(frame_list[i])
else:
if len(return_list) < len(tmp_list):
return_list = tmp_list
tmp_list = []
tmp_list.append(frame_list[i])
if len(return_list) < len(tmp_list):
return_list = tmp_list
return return_list
if __name__ == '__main__':
recognize_box("3.jpg", 245)
对图片位置及层级关系进行识别：（像素点）
import requests
import base64
from PIL import Image
import numpy as np
import json
def get_first_dot(image):
"""
获取颜色的分布 0是黑色 255 白色
为了扩展性更好，通过一行的长度最好
:param image:
:return:
"""
pixel_dict = []
rows, cols = image.size
for i in range(cols):
n = 0
color_list = []
color_max = []
for j in range(rows):
if image.getpixel((j, i)) > 0 and image.getpixel((j, i)) < 200:
color_list.append((j, i))
color_max.append(image.getpixel((j, i)))
# max = sorted(dict(zip(*np.unique(color_max, return_counts=True))).items(), key=lambda x:x[1])
# print("xx", max)
if len(color_list) > 250:
# print(color_list)
max_count = sorted(dict(zip(*np.unique(color_max, return_counts=True))).items(), key=lambda x: x[1])
# print(max)
# print(max[len(max) - 1:][0])
max_pixel_list = []
for m in color_list:
if image.getpixel(m) == max_count[len(max_count) - 1:][0][0]:
max_pixel_list.append(m)
# print(max_pixel_list)
first_location = sorted(max_pixel_list, key=lambda x: x[0])[0]
pixel_dict.append(first_location)
# print(pixel_dict)
result = []
for n in range(len(pixel_dict) - 1):
if len(result) == 0:
result.append(pixel_dict[n])
if pixel_dict[n + 1][1] - pixel_dict[n][1] > 20:
result.append(pixel_dict[n + 1])
print(result)
return result
def del_threshold_noise(image, threshold):
"""
直接通过阈值去除浅色点(作为第一步)，也要先转化为灰度
:return:
"""
pixel_matrix = image.load() # load之后可以直接操作，相当于转化为数组了
rows, cols = image.size
for col in range(0, cols):
for row in range(0, rows):
if pixel_matrix[row, col] >= threshold:
pixel_matrix[row, col] = 255
# image.save("1.jpg")
return image
def write_txt(dot_positions, data):
"""
将层级关系写入txt
:return:
"""
with open('1.txt', 'w') as f:
for index, value in enumerate(dot_positions):
if index == 0:
f.write(data[index] + "\n")
if index >= len(data) - 2:
pass
else:
if dot_positions[index + 1][0] - dot_positions[index][0] > 0:
f.write(data[index + 1] + "--" + "--" + "\n")
else:
f.write("--" + "--" + data[index + 1] + "\n")
if __name__ == '__main__':
img_path = '企业股权关系.jpg'
im = Image.open(img_path)
im = im.convert('L')
# get_threshold(im)
# 通过阈值先将杂点转化为白点，去除干扰； 范围：200-220
im = del_threshold_noise(im, 200)
dot_positions = get_first_dot(im)
write_txt(dot_positions, "123")
对验证码处理：
处理之前：
处理之后：
因为验证码干扰较多，处理之后识别的精度大概为75%左右，下面为处理过程：
import io
import math
import os
import cv2
import numpy as np
from PIL import Image
from collections import defaultdict
def del_threshold_noise(image, threshold):
"""
直接通过阈值去除浅色点(作为第一步)，也要先转化为灰度
:return:
"""
pixel_matrix = image.load() # load之后可以直接操作，相当于转化为数组了
rows, cols = image.size
for col in range(0, cols):
for row in range(0, rows):
if pixel_matrix[row, col] >= threshold:
pixel_matrix[row, col] = 255
return image
def get_threshold(image):
"""
获取像素点最多的像素(先转化为灰度)，就是二值化
:param image:
:return:
"""
pixel_dict = defaultdict(int)
rows, cols = image.size
for i in range(rows):
for j in range(cols):
pixel = image.getpixel((i, j))
pixel_dict[pixel] += 1
count_max = max(pixel_dict.values())
pixel_dict_reverse = {v: k for k, v in pixel_dict.items()}
threshold = pixel_dict_reverse[count_max]
return threshold
def get_bin_table(threshold):
"""
按照阈值进行二值化(先转化为灰度后，再进行二值化)
:param threshold: 像素阈值
:return:
"""
table = []
for i in range(256): # 0~256
rate = 0.1
if threshold * (1 - rate) <= i <= threshold * (1 + rate):
table.append(1) # 白色
else:
table.append(0) # 黑色
# for i in range(256): # 或者不做判断，直接找个界限（只找出黑色内容即可）
# if i < threshold:
# table.append(0)
# else:
# table.append(1)
return table
def del_cut_noise(im_cut):
'''
通过颜色区分：将图切为小图，找第二多颜色的像素，从而去除干扰线（先转化为灰度，入参和出参都是ndarray格式）
variable：bins：灰度直方图bin的数目
num_gray:像素间隔
method：1.找到灰度直方图中像素第二多所对应的像素
2.计算mode
3.除了在mode+-一定范围内的，全部变为空白。
'''
bins = 16 # 直方图柱子的数量，每个柱子都有一定的范围
num_gray = math.ceil(256 / bins) # 像素间隔就是柱子的范围
hist = cv2.calcHist([im_cut], [0], None, [bins], [0, 256])
lists = []
for i in range(len(hist)):
# print hist[i][0]
lists.append(hist[i][0])
second_max = sorted(lists)[-2] # 第二多的像素
bins_second_max = lists.index(second_max) # 第二多的像素是第几个柱子
mode = (bins_second_max + 0.5) * num_gray # 取柱子的中间（平均），比如2.5， 总的结果就是：第二多的平均的像素
for i in range(len(im_cut)):
for j in range(len(im_cut[0])):
if im_cut[i][j] < mode - 20 or im_cut[i][j] > mode + 20: # 数组可以直接操作
# print im_cut[i][j]
im_cut[i][j] = 255
return im_cut
def del_dot_noise(image):
"""
干扰点降噪
:param image:
:return:
"""
rows, cols = image.size # 图片的宽度和高度
change_pos = [] # 记录噪声点位置
# 遍历图片中的每个点，除掉边缘
for i in range(1, rows - 1):
for j in range(1, cols - 1):
# pixel_set用来记录该店附近的黑色像素的数量
pixel_set = []
# 取该点的邻域为以该点为中心的九宫格
for m in range(i - 1, i + 2):
for n in range(j - 1, j + 2):
if image.getpixel((m, n)) != 1: # 1为白色,0位黑色
pixel_set.append(image.getpixel((m, n)))
# 如果该位置的九宫内的黑色数量小于等于4，则判断为噪声
if len(pixel_set) <= 4:
change_pos.append((i, j))
# 对相应位置进行像素修改，将噪声处的像素置为1（白色）
for pos in change_pos:
image.putpixel(pos, 1) # 找到之后一起删除，而不是一个个删除
return image
def remove_noise_line(image):
"""
去除验证码干扰线（操作像素点：随机应变）
:param image:
:return:
"""
try:
width, height = image.size
total_list = []
for i in range(width):
dot_list = []
noise_dot_list = []
for j in range(height):
if image.getpixel((i, j)) < 200:
dot_list.append((i, j))
if i == 0:
if len(dot_list) == 1:
total_list.append(dot_list[0])
max = dot_list[0][1]
min = dot_list[0][1]
elif len(dot_list) == 2:
if dot_list[1][1] == dot_list[0][1] + 1:
total_list.append(dot_list[0])
total_list.append(dot_list[1])
max = dot_list[1][1]
min = dot_list[0][1]
elif abs(dot_list[0][1] - dot_list[1][1]) == 2:
total_list.append(dot_list[0])
total_list.append(dot_list[1])
max = dot_list[1][1]
min = dot_list[0][1]
elif len(dot_list) == 3:
if dot_list[1][1] == dot_list[0][1] + 1 and dot_list[2][1] == dot_list[0][1] + 2:
total_list.append(dot_list[0])
total_list.append(dot_list[1])
total_list.append(dot_list[2])
max = dot_list[2][1]
min = dot_list[0][1]
for m in dot_list:
if m[1] in range(min - 1, max + 2):
noise_dot_list.append(m)
# if max + 2 - min > 8:
# if m[1] in range(total_list[-2][1]-5, total_list[-2][1]+6):
# noise_dot_list.append(m)
# else:
# if m[1] in range(min-1, max+2):
# noise_dot_list.append(m)
noise_dot_list1 = []
noise_dot_list2 = []
# print('noise_dot_list', noise_dot_list)
if noise_dot_list:
if len(noise_dot_list) == noise_dot_list[-1][1] - noise_dot_list[0][1] + 1:
pass
else:
for index, value in enumerate(noise_dot_list):
if index > len(noise_dot_list) - 2:
break
if index == 0:
if value[1] + 1 == noise_dot_list[index + 1][1] and value[1] + 2 != noise_dot_list[index + 2][1]:
noise_dot_list1.append(value) if value not in noise_dot_list1 else 1
noise_dot_list1.append(noise_dot_list[index + 1]) if noise_dot_list[index + 1] not in noise_dot_list1 else 1
elif index == len(noise_dot_list) - 2:
if value[1] + 1 == noise_dot_list[index + 1][1]:
noise_dot_list1.append(value) if value not in noise_dot_list1 else 1
noise_dot_list1.append(noise_dot_list[index + 1]) if noise_dot_list[index + 1] not in noise_dot_list1 else 1
else:
if value[1] + 1 == noise_dot_list[index + 1][1] and value[1] + 2 != noise_dot_list[index + 2][1] and value[1] - 1 != noise_dot_list[index - 1][1]:
noise_dot_list1.append(value) if value not in noise_dot_list1 else 1
noise_dot_list1.append(noise_dot_list[index + 1]) if noise_dot_list[index + 1] not in noise_dot_list1 else 1
for index, value in enumerate(noise_dot_list):
if index > len(noise_dot_list) - 3:
break
if index == 0:
if value[1] + 1 == noise_dot_list[index + 1][1] and value[1] + 2 == noise_dot_list[index + 2][1] and value[1] + 3 != noise_dot_list[index + 3][1]:
noise_dot_list1.append(value) if value not in noise_dot_list1 else 1
noise_dot_list1.append(noise_dot_list[index + 1]) if noise_dot_list[index + 1] not in noise_dot_list1 else 1
noise_dot_list1.append(noise_dot_list[index + 2]) if noise_dot_list[index + 2] not in noise_dot_list1 else 1
elif index == len(noise_dot_list) - 3:
if value[1] + 1 == noise_dot_list[index + 1][1] and value[1] + 2 == noise_dot_list[index + 2][1] and value[1] - 1 != noise_dot_list[index - 1][1]:
noise_dot_list1.append(value) if value not in noise_dot_list1 else 1
noise_dot_list1.append(noise_dot_list[index + 1]) if noise_dot_list[index + 1] not in noise_dot_list1 else 1
noise_dot_list1.append(noise_dot_list[index + 2]) if noise_dot_list[index + 2] not in noise_dot_list1 else 1
else:
if value[1] + 1 == noise_dot_list[index + 1][1] and value[1] + 2 == noise_dot_list[index + 2][1] and value[1] + 3 != noise_dot_list[
index + 3][1] and value[1] - 1 != noise_dot_list[index - 1][1]:
noise_dot_list1.append(value) if value not in noise_dot_list1 else 1
noise_dot_list1.append(noise_dot_list[index + 1]) if noise_dot_list[index + 1] not in noise_dot_list1 else 1
noise_dot_list1.append(noise_dot_list[index + 2]) if noise_dot_list[index + 2] not in noise_dot_list1 else 1
# 找最近的两个或者三个
# print('total_list', total_list)
if noise_dot_list1:
d_value = sorted([abs(total_list[-2][1] - l[1]) for l in noise_dot_list1])[0]
mark = sorted([(total_list[-2][1] - l[1]) for l in noise_dot_list1])
if d_value in mark:
# print(total_list[-2][1] - d_value - 2)
# print(total_list[-2][1] -d_value + 3)
for i in noise_dot_list1:
if i[1] in range(total_list[-2][1] - d_value - 2, total_list[-2][1] -d_value + 3):
noise_dot_list2.append(i)
else:
# print(total_list[-2][1] + d_value - 2)
# print(d_value + total_list[-2][1] + 3)
for i in noise_dot_list1:
if i[1] in range(total_list[-2][1] + d_value - 2, d_value + total_list[-2][1] + 3):
noise_dot_list2.append(i)
# print('noise_dot_list1', noise_dot_list1)
# print('noise_dot_list2', noise_dot_list2)
if not noise_dot_list2:
count = 0
if noise_dot_list[0][1] != 0 and noise_dot_list[-1][1] != 39:
if image.getpixel((noise_dot_list[0][0], noise_dot_list[0][1] - 1)) < 200:
count += 1
if image.getpixel((noise_dot_list[-1][0], noise_dot_list[-1][1] + 1)) < 200:
count += 1
if len(noise_dot_list) + count < 4:
for n in noise_dot_list:
total_list.append(n)
else:
for n in noise_dot_list2:
total_list.append(n)
min = noise_dot_list[0][1]
max = noise_dot_list[-1][1]
else:
pass
# print(total_list)
for pos in total_list:
image.putpixel(pos, 255)
except Exception as e:
print(e)
return image
if __name__ == '__main__':
# 顺序：先灰度 --- 二值化 --- 降噪
# （如果通过点或者线降噪，降噪可以在二值化后，如果通过颜色降噪，要在二值化之前）
for i in os.listdir('test_images/'):
print(i)
im = Image.open('test_images/' + i)
# im = Image.open(io.BytesIO(f)) # f位读取的二进制
im = im.convert('L')
im = del_threshold_noise(im, 120)
# table = get_bin_table(240)
# out = im.point(table, '1')
# out = del_dot_noise(out)
out = remove_noise_line(im)
out.save('image_store/' + i)
# im = Image.open('test_images/1566286922439.png')
# im = im.convert('L')
# out = del_threshold_noise(im, 120)
# out.save('777.jpg')
# out = remove_noise_line(im)
# out.save('666.jpg')
扫一扫
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
