How to extract rectangular patches from a rotated rectangle?

Hi everyone,

I have to extract LBP features from a rotated rectangle section of an image. Since the extractLBPFeatures function takes as an input a NxM matrix, I thought about rotating my image to get an axis aligned rectangle. In the case of the image below I still get slanted edges because parts of the rectangle are out of the image.

So I though about creating small axis aligned rectangular patches to cover the maximum possible area of the rotate rectangle. I wasn't able to find such an alghoritm. Do you know any implementations?

Do you have any other idea how to extract the LBP features from the rotated rectangle?

Thank you in advance!

M

Answer 1

[Kuifeng]

 % top figure
% 1. edge finding, find the two parallel long lines
% 2. use polyfit to fit the slope of the longer line
% 3. draw the lines passing the top corner perpendicular to long line
% note Slope_longEdge * Slope_shortEdge = -1;

[martaR]

Thank you for your answer.

My problem is that the extractLBPFeatures function accept as an input the grayscale values of pixels in a matrix form (axis aligned MxN).

If I keep the rectangle slanted as you're suggesting I would have to input something like pixels:

(1,1)

(2,1) (2,2)

(3,1) (3,2) (3,3)

if my edge is rotated by 45 degrees. This is not possible.

Answer 2

[Image Analyst]

Attached is my demo of local binary pattern of an image. Adapt as needed.