Python cv2.getPerspectiveTransform() Examples

def Perspective_aug(src,strength,label=None):
    image = src
    pts_base = np.float32([[0, 0], [300, 0], [0, 300], [300, 300]])
    pts1=np.random.rand(4, 2)*random.uniform(-strength,strength)+pts_base
    pts1=pts1.astype(np.float32)
    #pts1 =np.float32([[56, 65], [368, 52], [28, 387], [389, 398]])
    M = cv2.getPerspectiveTransform(pts1, pts_base)
    trans_img = cv2.warpPerspective(image, M, (src.shape[1], src.shape[0]))

    label_rotated=None
    if label is not  None:
        label=label.T
        full_label = np.row_stack((label, np.ones(shape=(1, label.shape[1]))))
        label_rotated = np.dot(M, full_label)
        label_rotated=label_rotated.astype(np.int32)
        label_rotated=label_rotated.T
    return trans_img,label_rotated 

def align(image, points):
    """
    :param image:
    :param points:
    :return: aligned image
    """
    # alignment
    origin_point = np.require(np.array(points).reshape((4, 2)), dtype=np.single)
    height = int(max(np.linalg.norm(origin_point[0] - origin_point[1]), np.linalg.norm(origin_point[2] - origin_point[3])))
    width = int(max(np.linalg.norm(origin_point[0] - origin_point[3]), np.linalg.norm(origin_point[1] - origin_point[2])))

    target_point = np.float32([[0, 0], [0, height], [width, height], [width, 0]])
    map_matrix = cv2.getPerspectiveTransform(origin_point, target_point)
    cols = width + 1
    rows = height + 1
    color = cv2.warpPerspective(image, map_matrix, (cols, rows))
    return color 

def shift_scale_rotate(img, angle, scale, dx, dy):
    height, width = img.shape[:2]

    cc = math.cos(angle/180*math.pi) * scale
    ss = math.sin(angle/180*math.pi) * scale
    rotate_matrix = np.array([[cc, -ss], [ss, cc]])

    box0 = np.array([[0, 0], [width, 0],  [width, height], [0, height], ])
    box1 = box0 - np.array([width/2, height/2])
    box1 = np.dot(box1, rotate_matrix.T) + np.array([width/2+dx*width, height/2+dy*height])

    box0 = box0.astype(np.float32)
    box1 = box1.astype(np.float32)
    mat = cv2.getPerspectiveTransform(box0, box1)
    img = cv2.warpPerspective(img, mat, (width, height),
                              flags=cv2.INTER_LINEAR,
                              borderMode=cv2.BORDER_REFLECT_101)

    return img 

def shift_scale_rotate(img, angle, scale, dx, dy):
    height, width = img.shape[:2]

    cc = math.cos(angle/180*math.pi) * scale
    ss = math.sin(angle/180*math.pi) * scale
    rotate_matrix = np.array([[cc, -ss], [ss, cc]])

    box0 = np.array([[0, 0], [width, 0],  [width, height], [0, height], ])
    box1 = box0 - np.array([width/2, height/2])
    box1 = np.dot(box1, rotate_matrix.T) + np.array([width/2+dx*width, height/2+dy*height])

    box0 = box0.astype(np.float32)
    box1 = box1.astype(np.float32)
    mat = cv2.getPerspectiveTransform(box0, box1)
    img = cv2.warpPerspective(img, mat, (width, height),
                              flags=cv2.INTER_LINEAR,
                              borderMode=cv2.BORDER_REFLECT_101)

    return img 

def do_horizontal_shear(image, mask, scale=0):
    height, width = image.shape[:2]
    dx = int(scale * width)

    box0 = np.array([[0, 0], [width, 0], [width, height], [0, height], ], np.float32)
    box1 = np.array([[+dx, 0], [width + dx, 0], [width - dx, height], [-dx, height], ], np.float32)

    box0 = box0.astype(np.float32)
    box1 = box1.astype(np.float32)
    mat = cv2.getPerspectiveTransform(box0, box1)

    image = cv2.warpPerspective(image, mat, (width, height), flags=cv2.INTER_LINEAR,
                                borderMode=cv2.BORDER_REFLECT_101, borderValue=(0, 0, 0,))
    mask = cv2.warpPerspective(mask, mat, (width, height), flags=cv2.INTER_NEAREST,
                               borderMode=cv2.BORDER_REFLECT_101, borderValue=(0, 0, 0,))
    # mask = (mask > 0.5).astype(np.float32)
    return image, mask 

def do_rotation_transform(image, mask, angle=0):
    height, width = image.shape[:2]
    cc = np.cos(angle / 180 * np.pi)
    ss = np.sin(angle / 180 * np.pi)
    rotate_matrix = np.array([[cc, -ss], [ss, cc]])

    box0 = np.array([[0, 0], [width, 0], [width, height], [0, height], ], np.float32)
    box1 = box0 - np.array([width / 2, height / 2])
    box1 = np.dot(box1, rotate_matrix.T) + np.array([width / 2, height / 2])

    box0 = box0.astype(np.float32)
    box1 = box1.astype(np.float32)
    mat = cv2.getPerspectiveTransform(box0, box1)

    image = cv2.warpPerspective(image, mat, (width, height), flags=cv2.INTER_LINEAR,
                                borderMode=cv2.BORDER_REFLECT_101,
                                borderValue=(0, 0, 0,))
    mask = cv2.warpPerspective(mask, mat, (width, height), flags=cv2.INTER_NEAREST,
                               borderMode=cv2.BORDER_REFLECT_101,
                               borderValue=(0, 0, 0,))
    # mask = (mask > 0.5).astype(np.float32)
    return image, mask 

def shift_scale_rotate(img, angle, scale, dx, dy):
    height, width = img.shape[:2]

    cc = math.cos(angle/180*math.pi) * scale
    ss = math.sin(angle/180*math.pi) * scale
    rotate_matrix = np.array([[cc, -ss], [ss, cc]])

    box0 = np.array([[0, 0], [width, 0],  [width, height], [0, height], ])
    box1 = box0 - np.array([width/2, height/2])
    box1 = np.dot(box1, rotate_matrix.T) + np.array([width/2+dx*width, height/2+dy*height])

    box0 = box0.astype(np.float32)
    box1 = box1.astype(np.float32)
    mat = cv2.getPerspectiveTransform(box0, box1)
    img = cv2.warpPerspective(img, mat, (width, height),
                              flags=cv2.INTER_LINEAR,
                              borderMode=cv2.BORDER_REFLECT_101)

    return img 

def pivot_stitch(img, wd):
    # Stitch the area in between
    D = stitch(img[:, 1280 - wd:1280], img[:, 1280:1280 + wd], sigma=15.0)

    # Warp backwards
    pt1 = np.dot(D['H'], [wd, 400, 1])
    pt3 = np.dot(D['H'], [wd, 800, 1])
    pt1 = pt1 / pt1[2]
    pt3 = pt3 / pt3[2]
    src = np.zeros((4, 2), np.float32)
    dst = np.zeros((4, 2), np.float32)
    src[0] = [0, 0]
    src[1] = pt1[:2]
    src[2] = [0, 1280]
    src[3] = pt3[:2]
    dst = np.array(src)
    dst[1] = [2 * wd - 1, 400]
    dst[3] = [2 * wd - 1, 800]

    result = np.copy(img)
    M = cv2.getPerspectiveTransform(src, dst)
    result[:, 1280 - wd:1280 +
           wd] = cv2.warpPerspective(D['res'], M, (2 * wd, 1280))
    result[:, 1280 - wd:1280 + wd] = D['res']
    return result 

def Perspective_aug(src,strength,label=None):
    image = src
    pts_base = np.float32([[0, 0], [300, 0], [0, 300], [300, 300]])
    pts1=np.random.rand(4, 2)*random.uniform(-strength,strength)+pts_base
    pts1=pts1.astype(np.float32)
    #pts1 =np.float32([[56, 65], [368, 52], [28, 387], [389, 398]])
    M = cv2.getPerspectiveTransform(pts1, pts_base)
    trans_img = cv2.warpPerspective(image, M, (src.shape[1], src.shape[0]))

    label_rotated=None
    if label is not  None:
        label=label.T
        full_label = np.row_stack((label, np.ones(shape=(1, label.shape[1]))))
        label_rotated = np.dot(M, full_label)
        label_rotated=label_rotated.astype(np.int32)
        label_rotated=label_rotated.T
    return trans_img,label_rotated 

def Perspective_aug(src,strength,label=None):
    image = src
    pts_base = np.float32([[0, 0], [300, 0], [0, 300], [300, 300]])
    pts1=np.random.rand(4, 2)*random.uniform(-strength,strength)+pts_base
    pts1=pts1.astype(np.float32)
    #pts1 =np.float32([[56, 65], [368, 52], [28, 387], [389, 398]])
    M = cv2.getPerspectiveTransform(pts1, pts_base)
    trans_img = cv2.warpPerspective(image, M, (src.shape[1], src.shape[0]))

    label_rotated=None
    if label is not  None:
        label=label.T
        full_label = np.row_stack((label, np.ones(shape=(1, label.shape[1]))))
        label_rotated = np.dot(M, full_label)
        label_rotated=label_rotated.astype(np.int32)
        label_rotated=label_rotated.T
    return trans_img,label_rotated 

def shift_scale_rotate(img, angle, scale, dx, dy, borderMode=cv2.BORDER_CONSTANT):
    height, width = img.shape[:2]

    cc = math.cos(angle / 180 * math.pi) * scale
    ss = math.sin(angle / 180 * math.pi) * scale
    rotate_matrix = np.array([[cc, -ss], [ss, cc]])

    box0 = np.array([[0, 0], [width, 0], [width, height], [0, height], ])
    box1 = box0 - np.array([width / 2, height / 2])
    box1 = np.dot(box1, rotate_matrix.T) + np.array([width / 2 + dx * width, height / 2 + dy * height])

    box0 = box0.astype(np.float32)
    box1 = box1.astype(np.float32)
    mat = cv2.getPerspectiveTransform(box0, box1)
    img = cv2.warpPerspective(img, mat, (width, height),
                              flags=cv2.INTER_NEAREST,
                              borderMode=borderMode)

    return img 

def crop_and_warp(img, crop_rect):
	"""Crops and warps a rectangular section from an image into a square of similar size."""

	# Rectangle described by top left, top right, bottom right and bottom left points
	top_left, top_right, bottom_right, bottom_left = crop_rect[0], crop_rect[1], crop_rect[2], crop_rect[3]

	# Explicitly set the data type to float32 or `getPerspectiveTransform` will throw an error
	src = np.array([top_left, top_right, bottom_right, bottom_left], dtype='float32')

	# Get the longest side in the rectangle
	side = max([
		distance_between(bottom_right, top_right),
		distance_between(top_left, bottom_left),
		distance_between(bottom_right, bottom_left),
		distance_between(top_left, top_right)
	])

	# Describe a square with side of the calculated length, this is the new perspective we want to warp to
	dst = np.array([[0, 0], [side - 1, 0], [side - 1, side - 1], [0, side - 1]], dtype='float32')

	# Gets the transformation matrix for skewing the image to fit a square by comparing the 4 before and after points
	m = cv2.getPerspectiveTransform(src, dst)

	# Performs the transformation on the original image
	return cv2.warpPerspective(img, m, (int(side), int(side))) 

def crop_image(src, points, dst_height=None):
    """
    Crop heat map with points.
    :param src: 8-bit single-channel image (map).
    :param points: [[x1, y1], [x2, y2], [x3, y3], [x4, y4]].
    :return: dst_heat_map: Cropped image. 8-bit single-channel image (map) of heat map.
             src_points: [[x1, y1], [x2, y2], [x3, y3], [x4, y4]].
             dst_points: [[x1, y1], [x2, y2], [x3, y3], [x4, y4]].
    """
    src_image = src.copy()
    src_points = np.float32(points)
    width = round((cal_distance(points[0], points[1]) + cal_distance(points[2], points[3])) / 2)
    height = round((cal_distance(points[1], points[2]) + cal_distance(points[3], points[0])) / 2)
    if dst_height is not None:
        ratio = dst_height / min(width, height)
        width = int(width * ratio)
        height = int(height * ratio)
    crop_points = np.float32([[0, 0], [width, 0], [width, height], [0, height]])
    perspective_mat = cv2.getPerspectiveTransform(src=src_points, dst=crop_points)
    dst_heat_map = cv2.warpPerspective(src_image, perspective_mat, (width, height),
                                       borderValue=0, borderMode=cv2.BORDER_CONSTANT)
    return dst_heat_map, src_points, crop_points 

def un_warping(box, src_points, crop_points):
    """
    Unwarp the character bounding boxes.
    :param box: The character bounding box.
    :param src_points: Points before crop.
    :param crop_points: Points after crop.
    :return: The character bounding boxes after unwarp.
    """
    perspective_mat = cv2.getPerspectiveTransform(src=crop_points, dst=src_points)
    new_box = list()
    for x, y in box:
        new_x = int((perspective_mat[0][0] * x + perspective_mat[0][1] * y + perspective_mat[0][2]) /
                    (perspective_mat[2][0] * x + perspective_mat[2][1] * y + perspective_mat[2][2]))
        new_y = int((perspective_mat[1][0] * x + perspective_mat[1][1] * y + perspective_mat[1][2]) /
                    (perspective_mat[2][0] * x + perspective_mat[2][1] * y + perspective_mat[2][2]))
        new_box.append([new_x, new_y])
    return new_box 

def perspective_transform(src, dst_shape, dst_points):
        """
        Perspective Transform
        :param src: Image to transform.
        :param dst_shape: Tuple of 2 intergers(rows and columns).
        :param dst_points: [[x1, y1], [x2, y2], [x3, y3], [x4, y4]].
        :return: Image after perspective transform.
        """
        img = src.copy()
        h, w = img.shape[:2]
        src_points = np.float32([[0, 0], [w, 0], [w, h], [0, h]])
        dst_points = np.float32(dst_points)
        perspective_mat = cv2.getPerspectiveTransform(src=src_points, dst=dst_points)
        dst = cv2.warpPerspective(img, perspective_mat, (dst_shape[1], dst_shape[0]),
                                  borderValue=0, borderMode=cv2.BORDER_CONSTANT)
        return dst 

def perspective_transform(img, points):
    """Applies a 4-point perspective transformation in `img` so that `points`
    are the new corners."""
    source = np.array(
        points,
        dtype="float32")

    dest = np.array([
        [TRANSF_SIZE, TRANSF_SIZE],
        [0, TRANSF_SIZE],
        [0, 0],
        [TRANSF_SIZE, 0]],
        dtype="float32")

    transf = cv2.getPerspectiveTransform(source, dest)
    warped = cv2.warpPerspective(img, transf, (TRANSF_SIZE, TRANSF_SIZE))
    return warped 

def per_trans(img,cols,rows,img_path_mod):
    #3 rotational transformations
    #transformation 1
    pts7 = np.float32([[2,3],[93,4],[5,90],[92,91]])
    pts8 = np.float32([[0,0],[96,0],[0,96],[96,96]])
    M8 = cv2.getPerspectiveTransform(pts7,pts8)
    dst8 = cv2.warpPerspective(img,M8,(96,96))
    cv2.imwrite(img_path_mod + '_pt1.jpg',dst8)
    #transformation 2
    pts9 = np.float32([[6,7],[89,8],[9,87],[85,88]])
    pts10 = np.float32([[0,0],[96,0],[0,96],[96,96]])
    M9 = cv2.getPerspectiveTransform(pts9,pts10)
    dst9 = cv2.warpPerspective(img,M9,(96,96))
    cv2.imwrite(img_path_mod + '_pt2.jpg',dst9)
    #transformation 3
    pts11 = np.float32([[10,11],[93,12],[13,82],[83,84]])
    pts12 = np.float32([[0,0],[96,0],[0,96],[96,96]])
    M10 = cv2.getPerspectiveTransform(pts11,pts12)
    dst10 = cv2.warpPerspective(img,M10,(96,96))
    cv2.imwrite(img_path_mod + '_pt3.jpg',dst10) 

def draw_augmented_overlay(pts_1, overlay_image, image):
    """Overlay the image 'overlay_image' onto the image 'image'"""

    # Define the squares of the overlay_image image to be drawn:
    pts_2 = np.float32([[0, 0], [overlay_image.shape[1], 0], [overlay_image.shape[1], overlay_image.shape[0]],
                        [0, overlay_image.shape[0]]])

    # Draw border to see the limits of the image:
    cv2.rectangle(overlay_image, (0, 0), (overlay_image.shape[1], overlay_image.shape[0]), (255, 255, 0), 10)

    # Create the transformation matrix:
    M = cv2.getPerspectiveTransform(pts_2, pts_1)

    # Transform the overlay_image image using the transformation matrix M:
    dst_image = cv2.warpPerspective(overlay_image, M, (image.shape[1], image.shape[0]))
    # cv2.imshow("dst_image", dst_image)

    # Create the mask:
    dst_image_gray = cv2.cvtColor(dst_image, cv2.COLOR_BGR2GRAY)
    ret, mask = cv2.threshold(dst_image_gray, 0, 255, cv2.THRESH_BINARY_INV)

    # Compute bitwise conjunction using the calculated mask:
    image_masked = cv2.bitwise_and(image, image, mask=mask)
    # cv2.imshow("image_masked", image_masked)

    # Add the two images to create the resulting image:
    result = cv2.add(dst_image, image_masked)
    return result


# Create the dictionary object and the parameters: 

def tranfun(self, image):
        img = trans_utils.getcvimage(image)
        h,w = img.shape[:2]
        org = np.array([[0,np.random.randint(0,self.maxv)],
                        [w,np.random.randint(0,self.maxv)],
                        [0,h-np.random.randint(0,self.maxv)],
                        [w,h-np.random.randint(0,self.maxv)]],np.float32)
        dst = np.array([[0, 0], [w, 0], [0, h], [w, h]], np.float32)
        M = cv2.getPerspectiveTransform(org,dst)
        res = cv2.warpPerspective(img,M,(w,h))
        return getpilimage(res) 

def tranfun(self, image_and_loc):
        image, left, top, right, bottom = image_and_loc
        w, h = image.size
        left = np.clip(left,0,w-1)
        right = np.clip(right,0,w-1)
        top = np.clip(top, 0, h-1)
        bottom = np.clip(bottom, 0, h-1)
        img = trans_utils.getcvimage(image)
        try:
            # global index
            res = getpilimage(img[top:bottom,left:right])
            # res.save('test_imgs/crop-debug-{}.jpg'.format(index))
            # index+=1
            return res
        except AttributeError as e:
            print('error')
            image.save('test_imgs/t.png')
            print( left, top, right, bottom)

        h = bottom - top
        w = right - left
        org = np.array([[left - np.random.randint(0, self.maxv_w), top + np.random.randint(-self.maxv_h, self.maxv_h//2)],
                        [right + np.random.randint(0, self.maxv_w), top + np.random.randint(-self.maxv_h, self.maxv_h//2)],
                        [left - np.random.randint(0, self.maxv_w), bottom - np.random.randint(-self.maxv_h, self.maxv_h//2)],
                        [right + np.random.randint(0, self.maxv_w), bottom - np.random.randint(-self.maxv_h, self.maxv_h//2)]], np.float32)
        dst = np.array([[0, 0], [w, 0], [0, h], [w, h]], np.float32)
        M = cv2.getPerspectiveTransform(org,dst)
        res = cv2.warpPerspective(img,M,(w,h))
        return getpilimage(res) 

def four_point_transform(image, pts):
    # obtain a consistent order of the points and unpack them
    # individually
    rect = order_points(pts)
    (tl, tr, br, bl) = rect

    # compute the width of the new image, which will be the
    # maximum distance between bottom-right and bottom-left
    # x-coordiates or the top-right and top-left x-coordinates
    widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
    widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
    maxWidth = max(int(widthA), int(widthB))

    # compute the height of the new image, which will be the
    # maximum distance between the top-right and bottom-right
    # y-coordinates or the top-left and bottom-left y-coordinates
    heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
    heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
    maxHeight = max(int(heightA), int(heightB))

    # now that we have the dimensions of the new image, construct
    # the set of destination points to obtain a "birds eye view",
    # (i.e. top-down view) of the image, again specifying points
    # in the top-left, top-right, bottom-right, and bottom-left
    # order
    dst = np.array([
        [0, 0],
        [maxWidth - 1, 0],
        [maxWidth - 1, maxHeight - 1],
        [0, maxHeight - 1]], dtype="float32")

    # compute the perspective transform matrix and then apply it
    M = cv2.getPerspectiveTransform(rect, dst)
    warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))

    # return the warped image
    return warped 

def add_substitute_quad(image, substitute_quad, dst):
 
    # dst (zeroed) and src points
    dst = _order_points(dst)
 
    (tl, tr, br, bl) = dst
    min_x = min(int(tl[0]), int(bl[0]))
    min_y = min(int(tl[1]), int(tr[1]))
 
    for point in dst:
        point[0] = point[0] - min_x
        point[1] = point[1] - min_y
 
    (max_width,max_height) = _max_width_height(dst)
    src = _topdown_points(max_width, max_height)
 
    # warp perspective (with white border)
    substitute_quad = cv2.resize(substitute_quad, (max_width,max_height))
 
    warped = np.zeros((max_height,max_width,3), np.uint8)
    warped[:,:,:] = 255
 
    matrix = cv2.getPerspectiveTransform(src, dst)
    cv2.warpPerspective(substitute_quad, matrix, (max_width,max_height), warped, borderMode=cv2.BORDER_TRANSPARENT)
 
    # add substitute quad
    image[min_y:min_y + max_height, min_x:min_x + max_width] = warped
 
    return image 

def get_topdown_quad(image, src):

    # src and dst points
    src = _order_points(src)

    (max_width,max_height) = _max_width_height(src)
    dst = _topdown_points(max_width, max_height)
 
    # warp perspective
    matrix = cv2.getPerspectiveTransform(src, dst)
    warped = cv2.warpPerspective(image, matrix, _max_width_height(src))

    return warped 

def get_rand_transform_matrix(image_size, d, batch_size):
    Ms = np.zeros((batch_size, 2, 8))

    for i in range(batch_size):
        tl_x = random.uniform(-d, d)     # Top left corner, top
        tl_y = random.uniform(-d, d)    # Top left corner, left
        bl_x = random.uniform(-d, d)  # Bot left corner, bot
        bl_y = random.uniform(-d, d)    # Bot left corner, left
        tr_x = random.uniform(-d, d)     # Top right corner, top
        tr_y = random.uniform(-d, d)   # Top right corner, right
        br_x = random.uniform(-d, d)  # Bot right corner, bot
        br_y = random.uniform(-d, d)   # Bot right corner, right

        rect = np.array([
            [tl_x, tl_y],
            [tr_x + image_size, tr_y],
            [br_x + image_size, br_y + image_size],
            [bl_x, bl_y +  image_size]], dtype = "float32")

        dst = np.array([
            [0, 0],
            [image_size, 0],
            [image_size, image_size],
            [0, image_size]], dtype = "float32")

        M = cv2.getPerspectiveTransform(rect, dst)
        M_inv = np.linalg.inv(M)
        Ms[i,0,:] = M_inv.flatten()[:8]
        Ms[i,1,:] = M.flatten()[:8]
    return Ms 

def randomShiftScaleRotate(image, mask,
                           shift_limit=(-0.0625, 0.0625),
                           scale_limit=(-0.1, 0.1),
                           rotate_limit=(-45, 45), aspect_limit=(0, 0),
                           borderMode=cv2.BORDER_CONSTANT, u=0.5):
    if np.random.random() < u:
        height, width, channel = image.shape

        angle = np.random.uniform(rotate_limit[0], rotate_limit[1])  # degree
        scale = np.random.uniform(1 + scale_limit[0], 1 + scale_limit[1])
        aspect = np.random.uniform(1 + aspect_limit[0], 1 + aspect_limit[1])
        sx = scale * aspect / (aspect ** 0.5)
        sy = scale / (aspect ** 0.5)
        dx = round(np.random.uniform(shift_limit[0], shift_limit[1]) * width)
        dy = round(np.random.uniform(shift_limit[0], shift_limit[1]) * height)

        cc = np.math.cos(angle / 180 * np.math.pi) * sx
        ss = np.math.sin(angle / 180 * np.math.pi) * sy
        rotate_matrix = np.array([[cc, -ss], [ss, cc]])

        box0 = np.array([[0, 0], [width, 0], [width, height], [0, height], ])
        box1 = box0 - np.array([width / 2, height / 2])
        box1 = np.dot(box1, rotate_matrix.T) + np.array([width / 2 + dx, height / 2 + dy])

        box0 = box0.astype(np.float32)
        box1 = box1.astype(np.float32)
        mat = cv2.getPerspectiveTransform(box0, box1)
        image = cv2.warpPerspective(image, mat, (width, height), flags=cv2.INTER_LINEAR, borderMode=borderMode,
                                    borderValue=(
                                        0, 0,
                                        0,))
        mask = cv2.warpPerspective(mask, mat, (width, height), flags=cv2.INTER_LINEAR, borderMode=borderMode,
                                   borderValue=(
                                       0, 0,
                                       0,))

    return image, mask 

def main():
    global countClicks, coordinates, copyimage

    cv2.resizeWindow(windowname, 700, 700)

    while (countClicks < 4):
        preseedKey = cv2.waitKey(1)
        cv2.imshow(windowname, image)

        if preseedKey & 0xFF == 27:
            break

    pointone = np.float32(
        [[coordinates[0], coordinates[1]],
         [coordinates[2], coordinates[3]],
         [coordinates[4], coordinates[5]],
         [coordinates[6], coordinates[7]]])
    pointtwo = np.float32([[0, 0], [300, 0], [0, 300], [300, 300]])

    perspective = cv2.getPerspectiveTransform(pointone, pointtwo)
    output = cv2.warpPerspective(copyimage, perspective, (310, 310))

    cv2.imshow("Output Image", output)
    cv2.waitKey(0)

    cv2.destroyAllWindows() 

def rotRandrom(img, factor, size):
    shape = size
    pts1 = np.float32(
        [[0, 0], [0, shape[0]], [shape[1], 0], [shape[1], shape[0]]])
    pts2 = np.float32([[r(factor), r(factor)], [r(factor), shape[0] - r(factor)], [shape[1] - r(factor),  r(factor)],
                       [shape[1] - r(factor), shape[0] - r(factor)]])
    M = cv2.getPerspectiveTransform(pts1, pts2)
    dst = cv2.warpPerspective(img, M, size)
    return dst 

def rot(img, angel, shape, max_angel):
    """ 使图像轻微的畸变

        img 输入图像
        factor 畸变的参数
        size 为图片的目标尺寸

    """
    size_o = [shape[1], shape[0]]

    size = (shape[1] + int(shape[0] *
                           cos((float(max_angel) / 180) * 3.14)), shape[0])

    interval = abs(int(sin((float(angel) / 180) * 3.14) * shape[0]))

    pts1 = np.float32(
        [[0, 0], [0, size_o[1]], [size_o[0], 0], [size_o[0], size_o[1]]])
    if(angel > 0):

        pts2 = np.float32([[interval, 0], [0, size[1]], [size[0], 0], [
                          size[0] - interval, size_o[1]]])
    else:
        pts2 = np.float32([[0, 0], [interval, size[1]], [
                          size[0] - interval, 0], [size[0], size_o[1]]])

    M = cv2.getPerspectiveTransform(pts1, pts2)
    dst = cv2.warpPerspective(img, M, size)

    return dst 

def warp_perspective(self, rect, grid):
        (tl, tr, br, bl) = rect
        widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
        widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))

        # ...and now for the height of our new image
        heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
        heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))

        # take the maximum of the width and height values to reach
        # our final dimensions
        maxWidth = max(int(widthA), int(widthB))
        maxHeight = max(int(heightA), int(heightB))

        # construct our destination points which will be used to
        # map the screen to a top-down, "birds eye" view
        dst = np.array([
            [0, 0],
            [maxWidth - 1, 0],
            [maxWidth - 1, maxHeight - 1],
            [0, maxHeight - 1]], dtype="float32")

        # calculate the perspective transform matrix and warp
        # the perspective to grab the screen
        M = cv2.getPerspectiveTransform(rect, dst)
        warp = cv2.warpPerspective(grid, M, (maxWidth, maxHeight))
        return self.make_it_square(warp) 

def __call__(self, img, mask=None):
        if random.random() < self.prob:
            height, width, channel = img.shape

            angle = random.uniform(-self.rotate_limit, self.rotate_limit)
            scale = random.uniform(1-self.scale_limit, 1+self.scale_limit)
            dx = round(random.uniform(-self.shift_limit, self.shift_limit)) * width
            dy = round(random.uniform(-self.shift_limit, self.shift_limit)) * height

            cc = math.cos(angle/180*math.pi) * scale
            ss = math.sin(angle/180*math.pi) * scale
            rotate_matrix = np.array([[cc, -ss], [ss, cc]])

            box0 = np.array([[0, 0], [width, 0],  [width, height], [0, height], ])
            box1 = box0 - np.array([width/2, height/2])
            box1 = np.dot(box1, rotate_matrix.T) + np.array([width/2+dx, height/2+dy])

            box0 = box0.astype(np.float32)
            box1 = box1.astype(np.float32)
            mat = cv2.getPerspectiveTransform(box0, box1)
            img = cv2.warpPerspective(img, mat, (width, height),
                                      flags=cv2.INTER_LINEAR,
                                      borderMode=cv2.BORDER_REFLECT_101)
            if mask is not None:
                mask = cv2.warpPerspective(mask, mat, (width, height),
                                           flags=cv2.INTER_LINEAR,
                                           borderMode=cv2.BORDER_REFLECT_101)

        return img, mask