CODEWITHSUNDEEP
pythonopencvimage-processingcomputer-vision
Sophia
Sophia

Reputation: 407

What does cv2.cornerHarris return?

I have been reading about cornerHarris in OpenCV. I read through the documentation but I am not sure what does this function return.

While reading through the examples, there is a statement written as:

dst = cv2.cornerHarris(gray,2,3,0.04)

img[dst>0.01*dst.max()]=[0,0,255]

I just cannot understand the second statement above, may be because I do not understand, what is cornerHarris actually returning.

I do sense, that there is some kind of threshold being applied, but I cannot explain.

Upvotes: 4

Views: 6967

Answers (2)

Ashish Johnson
Ashish Johnson

Reputation: 467

dst = cv2.cornerHarris(gray,2,3,0.04)

dst is resultant 2D probability array with the same shape as of input image.

img[dst>0.01*dst.max()]=[0,0,255]

The above code is marely a post-processing technique, where "low-confidence-corners" are filtered out.

or to interpret differently:

Filter in those probabilities whose confidence is more than 1% of the max. probability in the resultant 2D probability array

From the examples documentation. If the image is less complex as given below, Even the below code will work fine:

img[dst == dst.max()]=[0,0,255]

Harris corner

For more clarity, plot the resultant 2D probability array using matplotlib:

def display(img, mapp='gray'):
    plt.imshow(img, cmap=mapp)
    plt.show()

Upvotes: 0

Sheradil
Sheradil

Reputation: 467

I just checked the example page you linked:

Threshold for an optimal value, it may vary depending on the image. img[dst>0.01*dst.max()]=[0,0,255]

This is the documentation page I was actually looking for: https://docs.opencv.org/2.4/modules/imgproc/doc/feature_detection.html?highlight=cornerharris#cornerharris

Corners in the image can be found as the local maxima of this response map.

The cv2.cornerHarris detector returns the corners that were found. As usual this method does not return a list of corners that were found. Instead, it returns an 2d-array - same size as the input - with probabilities. Each position inside this array holds the confidence in this pixel being a corner (to be precise it is not the single pixel, that is predicted to be a corner. It is the neighbourhood that is centered at this coordinates)

So most likely cv2.cornerHarris returns a value for confidence in it's own prediction. And with dst>0.01*dst.max() the code filters corners that are only "low-confidence-corners". Any corner that is detected will just be marked as a corner if the confidence is higher than 1% of the highest confidence.

Upvotes: 5

Related Questions