How to get the pixel coordinates of the region of interest using open-cv?

Question

I'm trying to get the pixel coordinates of a specific roi in a image. I created the roi using mask. The code and the result is shown below.

import cv2
import numpy as np

img = cv2.imread("Inxee.jpg")
img = cv2.resize(img, (640, 480))
mask = np.zeros(img.shape, np.uint8)

points  = np.array([[273,167], [363, 167], [573, 353], [63, 353]])  ##taking random points for ROI.
cv2.fillPoly(mask, [points], (100, 0, 100))

img = cv2.addWeighted(img, 0.7, mask, 0.5, 0)

values = img[np.where((mask == (100, 0, 100)).all(axis=1))]
print(values)

##cv2.imshow("mask", mask)
cv2.imshow("img", img)
cv2.waitKey(0)
cv2.destroyAllWindows()

result image

so in the image we can see the ROI.

I tried to use the

values = img[np.where((mask == (100, 0, 100)).all(axis=1))]

but here I'm getting only values not coordinates. So is there any way to get those coordinates?

you already have the corners of that polygon. you drew it yourself. you've presented an _approach_ towards something, but what is the _goal_? — Christoph Rackwitz, Mar 11 '22 at 12:24
the line you tried to use... can you figure out what the parts of that expression do? one part of it deals with indices/coordinates. — Christoph Rackwitz, Mar 11 '22 at 12:40
@ChristophRackwitz my goal is to fetch the coordinate of the each pixels within the ROI. — Shibaditya99, Mar 11 '22 at 12:47
@ChristophRackwitz Yes, within the image, where the color of the mask is (100, 0, 100), It's giving me the value of that pixel. So if I can find the coordinate from it then it's also preferable — Shibaditya99, Mar 11 '22 at 12:51
sure but what would you do with all these coordinates of pixels in the mask? knowing that is important to helping you. likely you don't actually need that data but something else. — Christoph Rackwitz, Mar 11 '22 at 13:03
yes I actually need that data, because i want to fuse LiDAR data with it and need too fetch all of the Z axis values according to the X axis value @ChristophRackwitz — Shibaditya99, Mar 12 '22 at 04:31

Shibaditya99 · Accepted Answer · 2022-03-12T09:15:13.073

2

Thanks for the solutions and possibilities friends, I just did,

val = np.where(mask < 0)
coordinate = list(zip(val[0], val[1]))
print(coordinate)

with this i got the coordinates! Thanks!

edited Mar 12 '22 at 09:15

answered Mar 12 '22 at 07:50

Shibaditya99

41
9

Olli · Answer 2 · 2022-03-11T14:22:58.090

To get the coordinates, you could do it like this:

pixels_in_roi = [(i_x,i_y) for i_x, col in enumerate(mask) for i_y, val in enumerate(col) if val == (100,0,100)]

There are faster ways to do it.

I'm not sure what your goal is in the end, but it sounds like undistorting this area as if it was a top-down view could be the next step. In that case this might help you: https://pyimagesearch.com/2014/08/25/4-point-opencv-getperspective-transform-example/

Edit: Here is a better solution using np.where: https://stackoverflow.com/a/27175491/7438122 The input 'mask' has to be a numpy array then (which it should already be in your case), not a list.

Christoph is right though: If you tell us what you want to achieve, there might be a way without extracting the indices at all.

yes, `np.where` is a faster way. I was trying to get OP to think about the code they have there, because it contains the solution already. -- since the question is unclear (what would one want with all the pixel _coordinates_ of a mask?), I _tried_ to get OP to explain. oh well. guessing is a tiring game on this site. — Christoph Rackwitz, Mar 11 '22 at 13:00

How to get the pixel coordinates of the region of interest using open-cv?

2 Answers2