HOG

Question 1

What is the purpose of HOG?

Answer 1

Using the image histogram to find similarities in images.

Question 2

What is PHOG, and how does it work?

Answer 2

PHOG is pyramid-HOG. You split the image into regions, and use HOG on all of those. So you might do HOG on the entire image first, then split into 4, do HOG. Split into 8 and so on. You use all the parts tho.

Question 3

What is image classification and how does it relate to HOG?

Answer 3

Image classification is detecting objects in images. You can use PHOG descriptors to find distances between histogram bins. This can be done on every level of the pyramid.

Question 4

How can you calculate the gradient in a color image?

Answer 4

• S-RGB method: Do for each channel and sum

- Max LAB: Convert to LAB colorspace and calculate gradient