Arguments to cv2::imshow

Question

Edit: original title "convert numpy array to cvmat" was a mistake - OpenCV's less than useful error message and my not reading the docs!

With OpenCV 2, IPython now uses NumPy arrays by default.

cvimage = cv2.imread("image.png") #using OpenCV 2
type(cvimage)
Out: numpy.ndarray  #dtype is uint8

pltimage = plt.imread("image.png")  #using Matplotlib
type(pltimage)
Out: numpy.ndarray   #dtype is float

plt.imshow(cvimage)  # works great

cv2.imshow(cvimage)
TypeError: Required argument 'mat' (pos 2) not found

Since cv2 uses NumPy arrays by default, there is no longer any cv::Mat constructor and NumPy has no functions to convert to a cv::Mat array.

Any ideas?

Answer 1

The function has the following docstring: imshow(winname, mat) -> None . You can see the doc string by typing cv2.imshow.__doc__ in the interpreter.

Try cv2.imshow('Image', cvimage) .

tl;dr : In original question, first argument of "window name" was missing. "imshow" takes two parameters and only one was supplied.

Answer 2

The question technically asks how to convert a NumPy Array (analogous to CV2 array) into a Mat object (CV). For anyone who is interested, this can be done by:

mat_array = cv.fromarray(numpy_array)

where mat_array is a Mat object, and numpy_array is a NumPy array or image. I would suggest staying away from older CV structures where possible. Numpy arrays offer much better performance than implemenations in native Python

Answer 3

Mat object was needed because C/C++ lacked a standard/native implementation of matrices.

However, numpy 's array is a perfect replacement for that functionality. Hence, the cv2 module accepts numpy.array s wherever a matrix is indicated in the docs.