Hog function from vlFeat

Hello guys,

I've got a question about HOG function from vlfeat. So, I use the function vl_hog to an 10*10 image with for example a cell size of 5pixels and number of bins 9. I should get in total 2 horizontal cells and 2 vertical cells, each cell containing 5 pixels.

Here comes my question, if I'm not mistaking the hog works like this: 5*5pixel = cell and 2*2cell = block. For each block we should get a feature vector of 4*9 = 36. Therefore, for my image I should get a feature vector of 36 values (because my image only contains 4 cells).

The problem is that the function vl_hog gives me as result a matrice output of (2,2,36). My question is why? Why it does not give me a single feature vector of 36 values. Where is my mistake.

These are the inputs when I use the function : numberOfBins = 9; cellsize = 5; image = 10*10; feav = vl_hog(image,cellsize,numberOfBins)

Can someone tell me where is my mistake pls? Why do I get a feature vector for each cell and not 1 single feature vector for the 4 cells (1 block)?

Here is the function : http://www.vlfeat.org/api/hog.html

Thank in advance you for your help!!