利用Python+OpenCV三步去除水印

时间：2022-09-17 10:28:36|栏目：Python代码|点击：次

一、推理原理

1.标定噪声的特征，使用cv2.inRange二值化标识噪声对图片进行二值化处理，具体代码：cv2.inRange(img, np.array([200, 200, 240]), np.array([255, 255, 255]))，把[200, 200, 200]~[255, 255, 255]以外的颜色处理为0

2.使用OpenCV的dilate方法，扩展特征的区域，优化图片处理效果

3.使用inpaint方法，把噪声的mask作为参数，推理并修复图片

二、推理步骤

1.从源图片，截取右下角部分，另存为新图片

2.识别水印，颜色值为：[200, 200, 200]~[255, 255, 255]

3.去掉水印，还原图片

4.把源图片、去掉水印的新图片，进行重叠合并

三、参考代码

import cv2
import numpy as np
from PIL import Image
import os

dir = os.getcwd()
path = "1.jpg"
newPath = "new.jpg"
img=cv2.imread(path,1)
hight,width,depth=img.shape[0:3]

#截取
cropped = img[int(hight*0.8):hight, int(width*0.7):width]  # 裁剪坐标为[y0:y1, x0:x1]
cv2.imwrite(newPath, cropped)
imgSY = cv2.imread(newPath,1)

#图片二值化处理，把[200,200,200]-[250,250,250]以外的颜色变成0
thresh = cv2.inRange(imgSY,np.array([200,200,200]),np.array([250,250,250]))
#创建形状和尺寸的结构元素
kernel = np.ones((3,3),np.uint8)
#扩展待修复区域
hi_mask = cv2.dilate(thresh,kernel,iterations=10)
specular = cv2.inpaint(imgSY,hi_mask,5,flags=cv2.INPAINT_TELEA)
cv2.imwrite(newPath, specular)

#覆盖图片
imgSY = Image.open(newPath)
img = Image.open(path)
img.paste(imgSY, (int(width*0.7),int(hight*0.8),width,hight))
img.save(newPath)

import cv2
import numpy as np
from PIL import Image
import os

dir = os.getcwd()
path = "1.jpg"
newPath = "new.jpg"
img=cv2.imread(path,1)
hight,width,depth=img.shape[0:3]

#截取
cropped = img[int(hight*0.8):hight, int(width*0.7):width]  # 裁剪坐标为[y0:y1, x0:x1]
cv2.imwrite(newPath, cropped)
imgSY = cv2.imread(newPath,1)

#图片二值化处理，把[200,200,200]-[250,250,250]以外的颜色变成0
thresh = cv2.inRange(imgSY,np.array([200,200,200]),np.array([250,250,250]))
#创建形状和尺寸的结构元素
kernel = np.ones((3,3),np.uint8)
#扩展待修复区域
hi_mask = cv2.dilate(thresh,kernel,iterations=10)
specular = cv2.inpaint(imgSY,hi_mask,5,flags=cv2.INPAINT_TELEA)
cv2.imwrite(newPath, specular)

#覆盖图片
imgSY = Image.open(newPath)
img = Image.open(path)
img.paste(imgSY, (int(width*0.7),int(hight*0.8),width,hight))
img.save(newPath)