Failure to use adaptiveThreshold: CV_8UC1 in function adaptiveThreshold

Question

I have used openCV python and encountered an error.

img_blur = cv2.medianBlur(self.cropped_img,5)
img_thresh_Gaussian = cv2.adaptiveThreshold(img_blur, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 11, 2)

plt.subplot(1,1,1),plt.imshow(img_thresh_Gaussian, cmap = 'gray')
plt.title("Image"), plt.xticks([]), plt.yticks([])
plt.show()

but I received:

cv2.error: /home/phuong/opencv_src/opencv/modules/imgproc/src/thresh.cpp:1280: error: (-215) src.type() == CV_8UC1 in function adaptiveThreshold

Do I have to install something else?

the input must be 8bit, single channel. you probably forgot to convert to grayscale even before applying medianblur. — berak, Nov 19 '14 at 10:18
CV_8UC1 is the source type whereas your `img_blur` type must be something different (i.e. need to be changed). The target needs to be either the same type i.e. CV8UC1 or bigger like CV16UC1 to cover for possible precision loss. I had similar issues in C++ API. — ha9u63a7, Nov 19 '14 at 10:18

score 149 · Answer 1 · answered Dec 27 '15 at 08:48

149

The problem is that you are trying to use adaptive thresholding to an image that is not in greyscale. And the function only works with a greyscale images.

So you have to convert your image to a greyscale format as it is described in documentation.

They read the image in a greyscale format with: img = cv2.imread('dave.jpg',0). You can also convert it to greyscale with: img_grey = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

answered Dec 27 '15 at 08:48

Salvador Dali

214,103
147
703
753

8

Excellent. I would take a good explanation over code correct any time... – Selam Getachew Jul 21 '17 at 23:19
7

I'd suggest using the named constant `cv2.IMREAD_GRAYSCALE` instead of the literal `0` as the second argument to `imread`. They're equal, but using the constant is significantly more self-documenting. – Mark Amery May 10 '18 at 12:14

score 16 · Accepted Answer · edited Feb 07 '16 at 15:45

16

you should load your file like this

src.create(rows, cols, CV_8UC1);
src = imread(your-file, CV_8UC1);

and after that

adaptiveThreshold(src, dst, 255, ADAPTIVE_THRESH_GAUSSIAN_C, THRESH_BINARY, 75, 10);

edited Feb 07 '16 at 15:45

David Schuler

2,375
1
14
6

answered Jan 26 '15 at 06:30

sam

1,363
1
20
32

9

what is ``src`` ``type`` – muon Mar 11 '17 at 22:48
14

-1; this only works by coincidence. The [`imread` docs](https://docs.opencv.org/master/d4/da8/group__imgcodecs.html#ga288b8b3da0892bd651fce07b3bbd3a56) dictate that `flags` passed to `imread` should be [`ImreadModes`](https://docs.opencv.org/master/d4/da8/group__imgcodecs.html#ga61d9b0126a3e57d9277ac48327799c80) - the constants that start with `IMREAD_`. The flag you pass, `CV_8UC1`, is not such a constant, and it's not particularly meaningful to use it as an `imread()` flag - but it happens to work because `CV_8UC1` and `IMREAD_GRAYSCALE` are both defined as `0`. – Mark Amery May 10 '18 at 12:13

score 13 · Answer 3 · answered Dec 01 '18 at 13:59

13

you can change the code to slightly like this :

img_blur = cv2.medianBlur(self.cropped_img,5).astype('uint8')
img_thresh_Gaussian = cv2.adaptiveThreshold(img_blur, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 11, 2)

just by adding ('uint8') while blurring has resolved my issue.

answered Dec 01 '18 at 13:59

Vaibhav K

2,762
3
21
22

This one solves my issue, thanks! – Zhuokai Zhao Sep 29 '21 at 04:20

score 6 · Answer 4 · edited May 10 '20 at 19:05

6

you have to imread your image in grayscale, so use the code below:

image = imread('address', 0)

0 exchange the channel of image to grayscale

edited May 10 '20 at 19:05

Adi Prasetyo

1,006
1
15
41

answered May 10 '20 at 11:10

Aram.rt

61
1
3

score 2 · Answer 5 · answered Mar 27 '20 at 11:17

2

Try to add cv2.IMREAD_GRAYSCALE in the second parameter of imread

img = cv2.imread(File_path, cv2.IMREAD_GRAYSCALE)

answered Mar 27 '20 at 11:17

Yazen WAKED

133
1
5

score 2 · Answer 6 · answered Sep 24 '20 at 06:59

Convert the image to grayscale before thresholding:

img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
img_blur = cv2.medianBlur(img_gray,5)
img_thresh_Gaussian = cv2.adaptiveThreshold(img_blur, 255,
        cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 11, 2)

score 1 · Answer 7 · answered Jan 03 '21 at 18:29

1

To solve this issue, you need to load grayscale image using 0(zero) as another parameter than the image path.
img = cv2.imread(imagepath, 0)
OR
img = cv2.imread(imagepath, cv2.IMREAD_GRAYSCALE)

answered Jan 03 '21 at 18:29

Vilas

51
4

I tried your solution and ones given by others on this thread. Nothing seems to be working. I am getting error on cv2.HoughLines(cdst, 1, np.pi / 180, 150, None, 0, 0) call with same error messages. – NewGuyInJava Jan 20 '22 at 16:23
Can you put your code and error here? @NewGuyInJava – Vilas Jan 29 '23 at 08:39

Failure to use adaptiveThreshold: CV_8UC1 in function adaptiveThreshold

7 Answers7

Linked