Hi,
As usual, a nice blog from AV. However, I’m a bit confused in one section where the histogram is created using gradient’s magnitude instead of the frequencies of the orientations within defined bins.
Section "Different Methods to Create Histograms using Gradients and Orientation’:
method 3:
As I understand, the histogram is something that shows the frequency of continuous data (single/univariate). But method 3, sounds like mapping between the magnitude and the orientation bins (relation between two variables). I’m not sure how would this be called a ‘Histogram’?
Further, I didn’t understand how is this realized? In the example given in the blog: Magnitude ‘13.6’ for a single pixel has an orientation of ‘36’ and so it falls in the bin (20-40). How would the table be when there are other pixels in the same image with different magnitudes having the same range of orientation? Say, there are 100 pixels of different magnitudes having orientations between 20 and 40? How does the table look like in this case?
EDIT: Sorry I missed out on another general question while posting. Adding it now. How are the edge cases (pixels on the border of the image) handled while calculating gradients? Is padding an option for this?
Thanks in advance
1 post - 1 participant