In OpenCV, cropping a rectangle from an image is extremely easy. However, it does not provide a function to crop a rotated rectangle. Here I developed an algorithm to crop a rotated rectangle from an image, and implemented it as a python function.
Original image and the rotated rectangle:
Crop a minimal up-right rectangle containing the rotated rectangle:
Rotate the cropped image such that the inner rotated rectangle becomes up-right and centered in the image:
Crop the "rotated rectangle":
- OpenCV 3.0
- Python 3.5
In Python or Jupyter Notebook:
from rotated_rect_crop import crop_rotated_rectangle
# image is the image matrix
# rect is the rotated rect object in OpenCV, i.e. (center (x,y), (width, height), angle of clock-wise rotation)
# center x is regarding to the width
# center y is regarding to the height
# width is the number of the columns of the rectangle.
# height is the number of the rows of the rectangle.
# The angle vector pointing to the "north" is exactly 0 degree
image_cropped = crop_rotated_rectangle(image, rect)Generate a random rotated rectangle and crop:
python rotated_rect_crop.py
