CODEWITHSUNDEEP
pythonopencvjupyter-notebook
Stefano Pozzi
Stefano Pozzi

Reputation: 619

COLOR_RGB2GRAY gives none-grayscale image (might be an issue with jupyter-notebook)

I am trying to perform histogram equalization on an image an have 2 issues. First, I need to plot the histogram for the grayscale version of it. When I try to convert the RGB image to grayscale the output is a blue and yellow image. My code goes as follows:

img = cv2.imread(r'D:/UNI/Y3/DIA/2K18/lab.jpg') 
RGB_img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
gray = cv2.cvtColor(RGB_img, cv2.COLOR_RGB2GRAY)
plt.imshow(gray)
plt.title('My picture (before hist. eq.)')
plt.show()

This is Jupyter Notebook's output:

This is Jupyter Notebook's output

But I just realized that if I save if it saves it correctly:

enter image description here

Since I need to hand in the jupyter doc how could I overcome this issue? Thx!

Second, I perform Histogram equalization but when trying to stack the images horizontally I get the following error form this code:

equ = cv2.equalizeHist(gray)
res = np.hstack((img,equ))

error -> all the input arrays must have same number of dimensions

As far as I can see I didn't touch the images' dimensions at all...

EDIT:

enter image description here

The left image should be RGB

Upvotes: 1

Views: 10838

Answers (1)

ZdaR
ZdaR

Reputation: 22954

As @Fredrik suggested, you can use plt.imshow(gray, cmap='gray', vmin = 0, vmax = 255) to get you a grayscale output or you can also convert the Grayscale image to a 3 channel RGB image using gray = cv2.cvtColor(gray, cv2.COLOR_GRAY2RGB).

Basically the problem is that gray = cv2.cvtColor(RGB_img, cv2.COLOR_RGB2GRAY) returns a single channel matrix, where as matplotlib is expecting a 3 channel RGB matrix to be shown, So you can either tell plt.imshow() that your input matrix is a single channel matrix using plt.imshow(gray, cmap='gray', vmin = 0, vmax = 255) or you can simply convert the single channel matrix to 3 channel matrix and then simply use plt.imshow(gray) and everything would work fine.

For the second part of your question where res = np.hstack((img,equ)) is raising error, it is always helpful to debug the shape of matrices, you want to apply operation upon, You can do that by print img.shape, print equ.shape. As far as I can see your img is a 3-channel matrix (BGR), whereas your equ is a single channel matrix (gray), hence error, you would need to convert this equ matrix to a 3-channel matrix again using cv2.cvtColor(equ, cv2.COLOR_GRAY2RGB).

Upvotes: 9

Related Questions