opencv+python用背景减去法抠图

opencv+python用背景减去法抠图

import cv2

from matplotlib import pyplot as plt

import numpy as np

imA=cv2.imread("target.png")

plt.imshow(cv2.cvtColor(imA,cv2.COLOR_BGR2RGB))

imA.shape #(2436, 1125, 3)

imBG=cv2.imread("bg_30061.jpg")

plt.imshow(cv2.cvtColor(imBG,cv2.COLOR_BGR2RGB))

# Step 1 预处理

# 日常缩放图像，背景图要缩放到和图A意一样大

imBG=cv2.resize(imBG,(1125,2436),interpolation=cv2.INTER_CUBIC)

# 后续代码都是基于灰度图像操作的

imA_gray=cv2.cvtColor(imA,cv2.COLOR_BGR2GRAY)

imBG_gray=cv2.cvtColor(imBG,cv2.COLOR_BGR2GRAY)

# 背景减去

sub=imBG_gray.astype("int32")-imA_gray.astype("int32")

sub=np.absolute(sub).astype("uint8")

plt.imshow(cv2.cvtColor(sub,cv2.COLOR_BGR2RGB))

# Step 2 二值化图像

thresh=cv2.adaptiveThreshold(sub,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C,cv2.THRESH_BINARY,19,5)

## 图像二值化有很多种方法， adaptiveThreshold 是基于局部的，最后两个参数，19,5 决定了边缘检测的清晰度

thresh=cv2.erode(thresh,None,iterations=2)

thresh=cv2.dilate(thresh,None,iterations=2)

# 膨胀腐蚀，主要是消去噪声点，同时边缘封闭(边缘不封闭，无法完全填充)

plt.figure(num="thresh")

plt.imshow(cv2.cvtColor(thresh,cv2.COLOR_BGR2RGB))

plt.close

# Step 3 边缘检测

cnts,hierarchy=cv2.findContours(thresh.copy(),cv2.RETR_TREE,cv2.CHAIN_APPROX_NONE)

len(cnts)

area=[]

for i in cnts:

cnt_area = cv2.contourArea(i)

area.append(cnt_area)

# 通过面积大小，找到需要的边缘的 index，这里为 866

mask=np.zeros([2436,1125],dtype=np.uint8)

mask[:,:]=255

res=cv2.drawContours(mask,cnts,866,(0,0,225),2)

plt.imshow(cv2.cvtColor(res,cv2.COLOR_BGR2RGB))

# 画出边缘

# 填充轮廓并制作掩模

mask2=np.zeros([2436,1125],dtype=np.uint8)

mask2[:,:]=0

res2=cv2.drawContours(mask2,cnts,866,255,cv2.FILLED)

plt.imshow(cv2.cvtColor(res2,cv2.COLOR_BGR2RGB))

# 保存为 png

h,w,c=imA.shape

b,g,r=cv2.split(imA)

imA_2=np.zeros((4,h,w),dtype=imA.dtype)

imA_2[0][0:h,0:w]=b

imA_2[1][0:h,0:w]=g

imA_2[2][0:h,0:w]=r

imA_2[3][0:h,0:w]=mask2

imA_new=cv2.merge(imA_2)

cv2.imwrite("imA_new.png",imA_new)

opencv+python用背景减去法抠图相关教程