Efficient way to threshold image array and render - Python / NumPy / OpenCV

Question

I have a camera that is sending the image data to my computer. From there my python script puts the 8bit color info (black and white; ranging from 0 - black - to 255 - white) into a numpy array. The array is 2D, first dimension up to 384 and second dimensions up to 288 Displaying this with a openCV window works great and the live video is more than 24fps.

My aim now is to manipulate the image, so that the live video displays any color value below 200 as 0 (completely black) and any color value above 200 as 255 (completely white). However, my code right now only gives me about 3fps.

My code is doing the following:

save the image that comes from the camera in a numpy array
open the first for loop to iterate through the x values
in that first for loop iterate through the second for loop with the y values
check each color value for every pixel and check if it is above 200 or not
depending on the if clause make the color value 0 or 255
display the new image

This is the decisive part in the code:

processedImage = frame
i = 0
for i in range(0, displayedWidth):
    ii = 0
    for ii in range(0, displayedHeight):
        if frame[ii, i] > 200:
            processedImage[ii, i] = 255
        else: 
            processedImage[ii, i] = 0
cv2.imshow("LiveVideo", processedImage)

I read here that for loops are faster than while loops, but it didn't improve the code's speed significantly, which is why I assume the rewriting of the processedImage takes too long.

Is there a way to make the whole process faster?

Thanks for any answers!

Seb · Accepted Answer · 2019-11-27 14:27:23Z

4

Try this:

frame[frame > 200] = 255
frame[frame <= 200] = 0
cv2.imshow("LiveVideo", frame)

answered Nov 27, 2019 at 14:27

Seb

4,60516 silver badges24 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

X_841 Over a year ago

Thank you very much! This made it go back to more than 24fps and the video is running smooth again. Will mark it as correct asap.

AKX · Accepted Answer · 2019-11-27 17:03:18Z

2

Staying with NumPy, it seems fastest would be -

(frame>200)*np.uint8(255)

Since, you are already using OpenCV, a faster way would be with cv2.threshold -

cv2.threshold(frame,200,255,cv2.THRESH_BINARY)[1]

Sample run to verify results and get timings on 1024X1024 image -

In [2]: np.random.seed(0)
   ...: frame = np.random.randint(0,256,(1024,1024)).astype(np.uint8)

In [3]: %timeit (frame>200)*np.uint8(255)
253 µs ± 13.7 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

In [4]: %timeit cv2.threshold(frame,200,255,cv2.THRESH_BINARY)[1]
58.2 µs ± 437 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)    

In [7]: out1 = (frame>200)*np.uint8(255)

In [8]: out2 = cv2.threshold(frame,200,255,cv2.THRESH_BINARY)[1]

In [9]: np.allclose(out1, out2)
Out[9]: True

Timings with other solutions on same data -

# @AKX's soln
In [10]: %timeit np.where(frame >= 200, 255, 0)
3.73 ms ± 22.7 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

# @Seb's soln
In [11]: %%timeit
    ...: frame[frame > 200] = 255
    ...: frame[frame <= 200] = 0
10.2 ms ± 15.4 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

OpenCV's version with cv2.threshold seems the fastest by a big margin among others.

edited Nov 27, 2019 at 17:03

AKX

171k16 gold badges147 silver badges229 bronze badges

answered Nov 27, 2019 at 15:03

Divakar

222k19 gold badges273 silver badges374 bronze badges

1 Comment

AwokeKnowing Over a year ago

i tried this ((frame>200)*np.uint8(255) ) and for my small one channel image it was even faster than cv2 thresh! so glad because that was the only non np thing i was doing

AKX · Accepted Answer · 2019-11-27 14:33:46Z

1

np.where() can be used too:

processed_frame = np.where(frame >= 200, 255, 0)

e.g.

>>> im = (np.random.rand(5, 5) * 255).round()
>>> im
array([[ 45.,  92.,  74., 207., 211.],
       [206., 184.,  31., 117., 119.],
       [ 88.,  82., 203.,  34.,   5.],
       [157.,  99., 154., 251.,  54.],
       [160., 177.,  14., 206., 234.]])
>>> np.where(im >= 200, 255, 0)
array([[  0,   0,   0, 255, 255],
       [255,   0,   0,   0,   0],
       [  0,   0, 255,   0,   0],
       [  0,   0,   0, 255,   0],
       [  0,   0,   0, 255, 255]])
>>>

answered Nov 27, 2019 at 14:33

AKX

171k16 gold badges147 silver badges229 bronze badges

Comments

foglerit · Accepted Answer · 2019-11-27 14:43:03Z

0

Using Seb's solution is a big improvement but it seems that using numpy.where is even faster:

%%timeit
frame1 = frame.copy()
frame1[frame1 > 200] = 255
frame1[frame1 <= 200] = 0

974 µs ± 19.1 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

vs.

%%timeit
frame2 = frame.copy()
frame2 = np.where(frame2 > 200, 255, 0)

385 µs ± 7.16 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

>>> np.alltrue(frame1 == frame2)
True

answered Nov 27, 2019 at 14:43

foglerit

8,4469 gold badges53 silver badges70 bronze badges

2 Comments

X_841 Over a year ago

When I try your or @AKX solution I get the error (-215: Assertion failed) when trying to display the image with cv2.imshow(...). Would I be able to save the points where the color value is above 200 with this as well, or would I need a second np.where() statement?

foglerit Over a year ago

Yes, the result should be the same as Seb's approach. For reference, I tested it with this array: frame = np.floor(256 * np.random.rand(384, 288)).astype(int)

Collectives™ on Stack Overflow

Efficient way to threshold image array and render - Python / NumPy / OpenCV

4 Answers 4

1 Comment

1 Comment

Comments

2 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

4 Answers 4

1 Comment

1 Comment

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related