在范围内对灰度图像进行阈值处理

Question

Does OpenCV cv.InRange function work only for RGB images?Can I do thresholding of grayscale image using this function?

I got an error,Following is my code:

   import cv2
   image=cv2.imread("disparitySGB.jpg")
   thresh=cv2.inRange(image,190,255);

It gives the following error:

thresh=cv2.inRange(image,190,255); TypeError: unknown is not a numpy array

I tried fixing it by:

  thresh=cv2.inRange(image,numpy.array(190),numpy.array(255));

Now there is no error but it produces black image.

Answer 1

For a gray-valued image which has shape (M, N) in numpy and size MxN with one single channel in OpenCV, then cv2.inRange takes scalar bounds:

gray = cv2.imread(filename, cv2.CV_LOAD_IMAGE_GRAYSCALE)
gray_filtered = cv2.inRange(gray, 190, 255)

But for RGB-images which have shape (M, N, 3) in numpy and size MxN with three channels in OpenCV you need to have the bounds match the "channel size".

rgb = cv2.imread(filename, cv2.CV_LOAD_IMAGE_COLOR)
rgb_filtered = cv2.inRange(gray, (190, 190, 190), (255, 255, 255))

This is explained in the documentation , although not very clearly.

Answer 2

cv2.inRange(src, lowerb, upperb[, dst]) → dst

Takes src as array and lower and upper as array or a scalar , this means you can use it to Threshold Grayscale images. You just have to use scalars for upper and lower .

Example:

myResult = cv2.InRange(myGrayscale, 50, 100)

Answer 3

你只需要'导入numpy as np'，你的原始代码应该可以正常工作。

Answer 4

Your cv2.imread is reading a RGB image. To read in grayscale it is

image = cv2.imread("disparitySGB.jpg", 0)