[英]How to mask out the object from the image?
I have a RGB image of 256X256 with the binary mask.我有一个带有二进制掩码的 256X256 的 RGB 图像。 I want to remove the object from the main image with the help of mask and keep the rest background.
我想借助掩码从主图像中删除 object 并保留 rest 背景。
I have tried with the below code but it didn't work.我已经尝试过使用下面的代码,但它没有用。
Can anyone let me know where I made the mistake?谁能让我知道我在哪里犯了错误?
import cv2
import numpy as np
# read image
img = cv2.imread('image.jpg')
mask2 = cv2.imread('mask.jpg') / 1
# mask by multiplication, clip to range 0 to 255 and make integer
result2 = (img * mask2).clip(0, 255).astype(np.uint8)
# save results
cv2.imwrite('result.png', result1)
Image, Mask:图片,面具:
Output I want, The result I am getting: Output 我想要,我得到的结果:
You need to use cv2.bitwise_and
for your problem.您需要使用
cv2.bitwise_and
来解决您的问题。 It performs 'AND' operation between your img
and mask
in a bitwise manner.它以按位方式在您的
img
和mask
之间执行“与”操作。
mask
, corresponding img
region is preservedmask
中的白色区域,保留对应的img
区域mask
, corresponding img
region is occludedmask
中的白色区域,对应的img
区域被遮挡Code:代码:
img = cv2.imread(r'C:\Users\524316\Desktop\Stack\eye\eye.jpg')
mask = cv2.imread(r'C:\Users\524316\Desktop\Stack\eye\mask.jpg', 0)
# invert the mask
th = cv2.threshold(mask,0,255,cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)[1]
# preserve the background
res = cv2.bitwise_and(img, img, mask = th)
Result:结果:
You don't need Otsu thresholding, just ordinary thresholding, with a sensible threshold.您不需要 Otsu 阈值,只需普通阈值,具有合理的阈值。 Your "mask" image is a JPEG.
您的“掩码”图像是 JPEG。 Lossy compression causes it to not be entirely black or white.
有损压缩会导致它不完全是黑色或白色。
Once you have the mask, you just need to learn about masking operations.有了遮罩后,您只需要了解遮罩操作即可。 That is a feature of
numpy
I'm using here:这是我在这里使用的
numpy
的一个功能:
im = cv.imread("VgAjK.jpg")
maskim = cv.imread("ZKKkp.jpg", cv.IMREAD_GRAYSCALE)
im[maskim >= 128] = 255 # "remove", paint over with white
You multiplied... that might have worked (except inverted) if your arrays had the right value ranges.如果您的 arrays 具有正确的值范围,那么您乘以...可能会起作用(倒置除外)。
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.