CODEWITHSUNDEEP
pythonimage-processingconvolutionopencv
Manusman
Manusman

Reputation: 57

Convolving an image with kernel returns empty image

I took on a challenge to implement a few OpenCV functions on my own in Python. Very many of these functions require convolving an image with a kernel, so this is the function I wrote to do that:

def convolve(img, kernel):  # img should have zero padding 
    kx, ky = kernel.shape  # kernel is square so kx == ky
    start, end = floor(kx / 2), ceil(kx / 2)
    x, y = img.shape
    convolved = np.zeros((x, y))

    for i in range(kx, x - kx):
        for j in range(kx, y - kx):
            convolved_area = img[i - start:i + end, j - start:j + end] * kernel
            convolved[i][j] = np.sum(convolved_area)

    return convolved

When I ran this on an image and a kernel that I made, I received a white image with the only color being the black padding.

For testing, the kernel that I used was a Gaussian kernel:

for x in range(radius):
    for y in range(radius):     # Generate kernel with formula
        kernel[x][y] = exp( -0.5 * (pow((x-mean)/sigma, 2.0) + pow((y-mean)/sigma,2.0)) ) / (2 * pi * sigma * sigma)
        magnitude += kernel[x][y]

for x in range(radius):     # normalize kernel
    for y in range(radius):
        kernel[x][y] /= magnitude

I know that the kernel works because using OpenCV's filter2D function worked for me:

convolved = cv2.filter2D(img, -1, kernel)

My question is why doesn't my convolve function work?

Upvotes: 1

Views: 460

Answers (1)

Red
Red

Reputation: 27547

You'll need to specify the datatype of uint8 for the image at

convolved = np.zeros((x, y))

making it

convolved = np.zeros((x, y), 'uint8')

But I would recommend using the np.zeros_like() method, which takes in an array and returns the array filled with zeros, preserving the original datatype of the array so that you won't have to specify uint8:

convolved = np.zeros_like(img)

The above is the most practical, but a few alternatives are

convolved = np.zeros((x, y)).astype('uint8')

and

convolved = np.uint8(np.zeros((x, y))

Upvotes: 1

Related Questions