Difference between cv2.NORM_L2 and cv2.NORM_L1 in opencv python

python opencv feature-detection
sp497 picture sp497 · Sep 29, 2015 · Viewed 9.3k times · Source

I am using sift algorithm from python extra modules for some feature matching. Although one thing I did not understand is the concept behind the normType passed to the BFMatcher. i.e Which ones have to be used in what case?

Any help will be invaluable

Answer

Miki picture Miki · Sep 29, 2015

From WolframAlpha NormL1 and NormL2:

Given a vector:

enter image description here

Norm L1 is the taxicab (or manhattan) distance (sum of absolute values):

enter image description here

while Norm L2 is the euclidean distance (square root of sum of squares):

enter image description here

The type of norm tells the BFMatcher how to compute the distances between every two features.

The NORM L1 is in general much faster to compute (mostly because you don't compute the square root). The NORM L2 is more accurate.

You can find a nice comparison here.

Related questions

OpenCV ORB detector finds very few keypoints

I'm trying to use the ORB keypoint detector and it seems to be returning much fewer points than the SIFT detector and the FAST detector. This image shows the keypoints found by the ORB detector: and this image shows the …

Read More
How to get pixel coordinates from Feature Matching in OpenCV Python

I need to get the list of the x and y coordinates of the pixels that the feature matcher selects in the code provided. I'm using Python and OpenCV. Can anyone help me? img1=cv2.imread('DSC_0216.jpg',0) img2=cv2.…

Read More
Detect touching/overlapping circles/ellipses with OpenCV and Python

i want to measure the circularity of circles (difference of the "circles" height and width or ellipse parameters). The circles are given in pictures as shown here: After doing usual stuff like color2gray, thresholding and border detection, I get …

Read More