How to do image processing inside box created by cv2.minAreaRect only?

Question

I draw a rotated rectangle in the original image. Is it possible to do image processing in the rotated rectangle box only without cropping it out from the original image?

original = cv2.imread(r'image.jpg')

ori_img = original.copy()
img = cv2.cvtColor(ori_img, cv2.COLOR_BGR2GRAY)
img = cv2.GaussianBlur(img,(5,5),0)
_, thresh = cv2.threshold(img, 130, 255, cv2.THRESH_BINARY_INV)
contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)

for cnt in contours:
    [x, y, w, h] = cv2.boundingRect(cnt)
    if (w*h) >= 10000:
        box = cv2.minAreaRect(np.asarray(cnt))
        pts = cv2.boxPoints(box)
        pts = np.int0(pts)
cv2.drawContours(ori_img, [pts], 0, (0,255,0), 3)    
cv2.namedWindow('Frame', cv2.WINDOW_NORMAL)
cv2.imshow('Frame',ori_img)
k = cv2.waitKey(0)

Answer 1

Short answer: no .

No OpenCV functions take a RotatedRect to limit the area they are working on.

Long answer: maybe .

Some OpenCV functions can take a mask argument. Many don't.

To calculate a mask from the rotated rectangle, use boxPoints and then draw a filled polygon with color 255 into an array of dtype uint8 and shape (height, width)

mask = np.zeros((height, width), dtype=np.uint8)

# pts from boxPoints is a float32 array, fillPoly only likes integers
cv.fillPoly(
    img=mask,
    pts=np.round(pts).astype(np.int32).reshape((-1, 1, 2)), # round, cast, reshape into 2-channel column vector
    color=255)

Longer answer: do your operations, then copy back using the mask.

source = ...
altered = your_operations(source)
source[mask != 0] = altered[mask != 0]

To reduce the amount of work, you can do your operations on a subregion too (and then use mask to copy back).