Python cv2.BORDER_REPLICATE Examples

def affine_skew(self, tilt, phi, img, mask=None):
        h, w = img.shape[:2]
        if mask is None:
            mask = np.zeros((h, w), np.uint8)
            mask[:] = 255
        A = np.float32([[1, 0, 0], [0, 1, 0]])
        if phi != 0.0:
            phi = np.deg2rad(phi)
            s, c = np.sin(phi), np.cos(phi)
            A = np.float32([[c, -s], [s, c]])
            corners = [[0, 0], [w, 0], [w, h], [0, h]]
            tcorners = np.int32(np.dot(corners, A.T))
            x, y, w, h = cv2.boundingRect(tcorners.reshape(1, -1, 2))
            A = np.hstack([A, [[-x], [-y]]])
            img = cv2.warpAffine(img, A, (w, h), flags=cv2.INTER_LINEAR, borderMode=cv2.BORDER_REPLICATE)
        if tilt != 1.0:
            s = 0.8*np.sqrt(tilt * tilt - 1)
            img = cv2.GaussianBlur(img, (0, 0), sigmaX=s, sigmaY=0.01)
            img = cv2.resize(img, (0, 0), fx=1.0 / tilt, fy=1.0, interpolation=cv2.INTER_NEAREST)
            A[0] /= tilt
        if phi != 0.0 or tilt != 1.0:
            h, w = img.shape[:2]
            mask = cv2.warpAffine(mask, A, (w, h), flags=cv2.INTER_NEAREST)
        Ai = cv2.invertAffineTransform(A)
        return img, mask, Ai 

def __init__(self, max_deg, center_range=(0, 1),
                 interp=cv2.INTER_LINEAR,
                 border=cv2.BORDER_REPLICATE, step_deg=None, border_value=0):
        """
        Args:
            max_deg (float): max abs value of the rotation angle (in degree).
            center_range (tuple): (min, max) range of the random rotation center.
            interp: cv2 interpolation method
            border: cv2 border method
            step_deg (float): if not None, the stepping of the rotation
                angle. The rotation angle will be a multiple of step_deg. This
                option requires ``max_deg==180`` and step_deg has to be a divisor of 180)
            border_value: cv2 border value for border=cv2.BORDER_CONSTANT
        """
        assert step_deg is None or (max_deg == 180 and max_deg % step_deg == 0)
        super(Rotation, self).__init__()
        self._init(locals()) 

def __init__(self, p_flip=0.0, max_rotation=0.0, max_shear=0.0, max_scale=0.0, max_ar_factor=0.0,
                 border_mode='constant', pad_amount=0):
        super().__init__()
        self.p_flip = p_flip
        self.max_rotation = max_rotation
        self.max_shear = max_shear
        self.max_scale = max_scale
        self.max_ar_factor = max_ar_factor

        if border_mode == 'constant':
            self.border_flag = cv.BORDER_CONSTANT
        elif border_mode == 'replicate':
            self.border_flag == cv.BORDER_REPLICATE
        else:
            raise Exception

        self.pad_amount = pad_amount 

def __init__(self, max_deg, center_range=(0, 1),
                 interp=cv2.INTER_LINEAR,
                 border=cv2.BORDER_REPLICATE, step_deg=None, border_value=0):
        """
        Args:
            max_deg (float): max abs value of the rotation angle (in degree).
            center_range (tuple): (min, max) range of the random rotation center.
            interp: cv2 interpolation method
            border: cv2 border method
            step_deg (float): if not None, the stepping of the rotation
                angle. The rotation angle will be a multiple of step_deg. This
                option requires ``max_deg==180`` and step_deg has to be a divisor of 180)
            border_value: cv2 border value for border=cv2.BORDER_CONSTANT
        """
        assert step_deg is None or (max_deg == 180 and max_deg % step_deg == 0)
        super(Rotation, self).__init__()
        self._init(locals()) 

def __init__(self, max_deg, center_range=(0, 1),
                 interp=cv2.INTER_LINEAR,
                 border=cv2.BORDER_REPLICATE, step_deg=None, border_value=0):
        """
        Args:
            max_deg (float): max abs value of the rotation angle (in degree).
            center_range (tuple): (min, max) range of the random rotation center.
            interp: cv2 interpolation method
            border: cv2 border method
            step_deg (float): if not None, the stepping of the rotation
                angle. The rotation angle will be a multiple of step_deg. This
                option requires ``max_deg==180`` and step_deg has to be a divisor of 180)
            border_value: cv2 border value for border=cv2.BORDER_CONSTANT
        """
        assert step_deg is None or (max_deg == 180 and max_deg % step_deg == 0)
        super(Rotation, self).__init__()
        self._init(locals()) 

def transform_state(t, **kwargs):
    if callable(t):
        t_vars = vars(t)

        if 'random_state' in kwargs and 'random' in t_vars:
            t.__dict__['random'] = kwargs['random_state']

        support = ['fillval', 'anchor', 'prob', 'mean', 'std', 'outside']
        for arg in kwargs:
            if arg in t_vars and arg in support:
                t.__dict__[arg] = kwargs[arg]

        if 'mode' in kwargs and 'mode' in t_vars:
            t.__dict__['mode'] = kwargs['mode']
        if 'border' in kwargs and 'border' in t_vars:
            t.__dict__['border'] = BorderTypes.get(kwargs['border'], cv2.BORDER_REPLICATE)

        if 'transforms' in t_vars:
            t.__dict__['transforms'] = transforms_state(t.transforms, **kwargs)
    return t 

def distort_affine_cv2(image, alpha_affine=10, random_state=None):
    if random_state is None:
        random_state = np.random.RandomState(None)

    shape = image.shape
    shape_size = shape[:2]

    center_square = np.float32(shape_size) // 2
    square_size = min(shape_size) // 3
    pts1 = np.float32([
        center_square + square_size,
        [center_square[0] + square_size, center_square[1] - square_size],
        center_square - square_size])
    pts2 = pts1 + random_state.uniform(-alpha_affine, alpha_affine, size=pts1.shape).astype(np.float32)

    M = cv2.getAffineTransform(pts1, pts2)
    distorted_image = cv2.warpAffine(
        image, M, shape_size[::-1], borderMode=cv2.BORDER_REPLICATE) #cv2.BORDER_REFLECT_101)

    return distorted_image 

def _crop(self, image, center, size):
        corners = np.zeros(4, dtype=int)
        corners[:2] = np.floor(center - size / 2).astype(int)
        corners[2:] = corners[:2] + size
        pads = np.concatenate(
            (-corners[:2], corners[2:] - image.shape[1::-1]))
        pads = np.maximum(0, pads)

        if np.any(pads > 0):
            corners = np.concatenate((
                corners[:2] + pads[:2],
                corners[2:] - pads[2:])).astype(int)

        patch = image[corners[1]:corners[3], corners[0]:corners[2]]

        if np.any(pads > 0):
            patch = cv2.copyMakeBorder(
                patch, pads[1], pads[3], pads[0], pads[2],
                borderType=cv2.BORDER_REPLICATE)

        return patch 

def _crop(self, image, center, size):
        corners = np.zeros(4, dtype=int)
        corners[:2] = np.floor(center - size / 2).astype(int)
        corners[2:] = corners[:2] + size
        pads = np.concatenate(
            (-corners[:2], corners[2:] - image.shape[1::-1]))
        pads = np.maximum(0, pads)

        if np.any(pads > 0):
            corners = np.concatenate((
                corners[:2] + pads[:2],
                corners[2:] - pads[2:])).astype(int)

        patch = image[corners[1]:corners[3], corners[0]:corners[2]]

        if np.any(pads > 0):
            patch = cv2.copyMakeBorder(
                patch, pads[1], pads[3], pads[0], pads[2],
                borderType=cv2.BORDER_REPLICATE)

        return patch 

def _crop(self, image, center, size):
        corners = np.zeros(4, dtype=int)
        corners[:2] = np.floor(center - size / 2).astype(int)
        corners[2:] = corners[:2] + size
        pads = np.concatenate(
            (-corners[:2], corners[2:] - image.shape[1::-1]))
        pads = np.maximum(0, pads)

        if np.any(pads > 0):
            corners = np.concatenate((
                corners[:2] + pads[:2],
                corners[2:] - pads[2:])).astype(int)

        patch = image[corners[1]:corners[3], corners[0]:corners[2]]

        if np.any(pads > 0):
            patch = cv2.copyMakeBorder(
                patch, pads[1], pads[3], pads[0], pads[2],
                borderType=cv2.BORDER_REPLICATE)

        return patch 

def _crop(self, image, center, size):
        corners = np.zeros(4, dtype=int)
        corners[:2] = np.floor(center - size / 2).astype(int)
        corners[2:] = corners[:2] + size
        pads = np.concatenate(
            (-corners[:2], corners[2:] - image.shape[1::-1]))
        pads = np.maximum(0, pads)

        if np.any(pads > 0):
            corners = np.concatenate((
                corners[:2] + pads[:2],
                corners[2:] - pads[2:])).astype(int)

        patch = image[corners[1]:corners[3], corners[0]:corners[2]]

        if np.any(pads > 0):
            patch = cv2.copyMakeBorder(
                patch, pads[1], pads[3], pads[0], pads[2],
                borderType=cv2.BORDER_REPLICATE)

        return patch 

def preprocess(image, width, height):
    # Grab the dimensions of the image, then initialize the padding values
    (h, w) = image.shape[:2]

    # If the width is greater than the height then resize along the width
    if w > h:
        image = imutils.resize(image, width=width)
    # Otherwise, the height is greater than the width so resize along the height
    else:
        image = imutils.resize(image, height=height)

    # Determine the padding values for the width and height to obtain the target dimensions
    pad_w = int((width - image.shape[1]) / 2.0)
    pad_h = int((height - image.shape[0]) / 2.0)

    # Pad the image then apply one more resizing to handle any rounding issues
    image = cv2.copyMakeBorder(image, pad_h, pad_h, pad_w, pad_w, cv2.BORDER_REPLICATE)
    image = cv2.resize(image, (width, height))

    # Return the pre-processed image
    return image 

def preprocess(image, width, height):
    # Grab the dimensions of the image, then initialize the padding values
    (h, w) = image.shape[:2]

    # If the width is greater than the height then resize along the width
    if w > h:
        image = imutils.resize(image, width=width)
    # Otherwise, the height is greater than the width so resize along the height
    else:
        image = imutils.resize(image, height=height)

    # Determine the padding values for the width and height to obtain the target dimensions
    pad_w = int((width - image.shape[1]) / 2.0)
    pad_h = int((height - image.shape[0]) / 2.0)

    # Pad the image then apply one more resizing to handle any rounding issues
    image = cv2.copyMakeBorder(image, pad_h, pad_h, pad_w, pad_w, cv2.BORDER_REPLICATE)
    image = cv2.resize(image, (width, height))

    # Return the pre-processed image
    return image 

def padImg(img):
    '''
    pad image twice.
    The first padding is to make sure the patches cover all image regions.
    The second padding is used for cropping the global patch.
    '''
    
    H = img.shape[0]
    W = img.shape[1]
    
    globalFct = 4
    patchRes = 256
    ovlp = int(patchRes * 0.25)
    
    padH = (int((H - patchRes)/(patchRes - ovlp) + 1) * (patchRes - ovlp) + patchRes) - H
    padW = (int((W - patchRes)/(patchRes - ovlp) + 1) * (patchRes - ovlp) + patchRes) - W
    
    padding = int(patchRes * (globalFct - 1) / 2.0)

    padImg = cv2.copyMakeBorder(img, 0, padH, 0, padW, cv2.BORDER_REPLICATE)
    padImg = cv2.copyMakeBorder(padImg, padding, padding, padding, padding, cv2.BORDER_REPLICATE)
    
    return padImg 

def __init__(self, tx=(-0.1, 0.1), ty=None, border='constant', fillval=0, anchor=None, random_state=np.random):   
        if isinstance(tx, numbers.Number):
            tx = (-abs(tx), abs(tx))
        assert isinstance(tx, tuple) and np.abs(tx).max() < 1
        if ty is None:
            ty = tx
        elif isinstance(ty, numbers.Number):
            ty = (-abs(ty), abs(ty))
        assert isinstance(ty, tuple) and np.abs(ty).max() < 1
        self.tx, self.ty = tx, ty

        if isinstance(fillval, numbers.Number):
            fillval = [fillval] * 3

        self.border = BorderTypes.get(border, cv2.BORDER_REPLICATE)
        self.fillval = fillval
        self.anchor = anchor
        self.random = random_state 

def _augment(self, img, s):
        return np.reshape(cv2.GaussianBlur(img, s, sigmaX=0, sigmaY=0,
                                           borderType=cv2.BORDER_REPLICATE), img.shape) 

def __init__(self, scale=None, translate_frac=None, rotate_max_deg=0.0, shear=0.0,
                 interp=cv2.INTER_LINEAR, border=cv2.BORDER_REPLICATE, border_value=0):
        """
        Args:
            scale (tuple of 2 floats): scaling factor interval, e.g (a, b), then scale is
                randomly sampled from the range a <= scale <= b. Will keep
                original scale by default.
            translate_frac (tuple of 2 floats): tuple of max abs fraction for horizontal
                and vertical translation. For example translate_frac=(a, b), then horizontal shift
                is randomly sampled in the range 0 < dx < img_width * a and vertical shift is
                randomly sampled in the range 0 < dy < img_height * b. Will
                not translate by default.
            shear (float): max abs shear value in degrees between 0 to 180
            interp: cv2 interpolation method
            border: cv2 border method
            border_value: cv2 border value for border=cv2.BORDER_CONSTANT
        """
        if scale is not None:
            assert isinstance(scale, tuple) and len(scale) == 2, \
                "Argument scale should be a tuple of two floats, e.g (a, b)"

        if translate_frac is not None:
            assert isinstance(translate_frac, tuple) and len(translate_frac) == 2, \
                "Argument translate_frac should be a tuple of two floats, e.g (a, b)"

        assert shear >= 0.0, "Argument shear should be between 0.0 and 180.0"

        super(Affine, self).__init__()
        self._init(locals()) 

def _align_face_projection(img, src_lmk, dst_lmk, size=None):
    H, _ = cv2.findHomography(src_lmk, dst_lmk)
    if size is None:
        size = (img.shape[1], img.shape[0])
    # TODO: Border method
    aligned_img = cv2.warpPerspective(img, H, size,
                                      borderMode=cv2.BORDER_REPLICATE)
    return aligned_img 

def __init__(self, scale=None, translate_frac=None, rotate_max_deg=0.0, shear=0.0,
                 interp=cv2.INTER_LINEAR, border=cv2.BORDER_REPLICATE, border_value=0):
        """
        Args:
            scale (tuple of 2 floats): scaling factor interval, e.g (a, b), then scale is
                randomly sampled from the range a <= scale <= b. Will keep
                original scale by default.
            translate_frac (tuple of 2 floats): tuple of max abs fraction for horizontal
                and vertical translation. For example translate_frac=(a, b), then horizontal shift
                is randomly sampled in the range 0 < dx < img_width * a and vertical shift is
                randomly sampled in the range 0 < dy < img_height * b. Will
                not translate by default.
            shear (float): max abs shear value in degrees between 0 to 180
            interp: cv2 interpolation method
            border: cv2 border method
            border_value: cv2 border value for border=cv2.BORDER_CONSTANT
        """
        if scale is not None:
            assert isinstance(scale, tuple) and len(scale) == 2, \
                "Argument scale should be a tuple of two floats, e.g (a, b)"

        if translate_frac is not None:
            assert isinstance(translate_frac, tuple) and len(translate_frac) == 2, \
                "Argument translate_frac should be a tuple of two floats, e.g (a, b)"

        assert shear >= 0.0, "Argument shear should be between 0.0 and 180.0"

        super(Affine, self).__init__()
        self._init(locals()) 

def rotate_image(self, img_patch, slope):
        (h, w) = img_patch.shape[:2]
        center = (w // 2, h // 2)
        M = cv2.getRotationMatrix2D(center, slope, 1.0)
        return cv2.warpAffine(img_patch, M, (w, h), flags=cv2.INTER_CUBIC, borderMode=cv2.BORDER_REPLICATE) 

def _sample_patch(self, im, pos, sample_sz, output_sz):
        pos = np.floor(pos)
        sample_sz = np.maximum(mround(sample_sz), 1)
        xs = np.floor(pos[1]) + np.arange(0, sample_sz[1]+1) - np.floor((sample_sz[1]+1)/2)
        ys = np.floor(pos[0]) + np.arange(0, sample_sz[0]+1) - np.floor((sample_sz[0]+1)/2)
        xmin = max(0, int(xs.min()))
        xmax = min(im.shape[1], int(xs.max()))
        ymin = max(0, int(ys.min()))
        ymax = min(im.shape[0], int(ys.max()))
        # extract image
        im_patch = im[ymin:ymax, xmin:xmax, :]
        left = right = top = down = 0
        if xs.min() < 0:
            left = int(abs(xs.min()))
        if xs.max() > im.shape[1]:
            right = int(xs.max() - im.shape[1])
        if ys.min() < 0:
            top = int(abs(ys.min()))
        if ys.max() > im.shape[0]:
            down = int(ys.max() - im.shape[0])
        if left != 0 or right != 0 or top != 0 or down != 0:
            im_patch = cv2.copyMakeBorder(im_patch, top, down, left, right, cv2.BORDER_REPLICATE)
        # im_patch = cv2.resize(im_patch, (int(output_sz[0]), int(output_sz[1])))
        im_patch = cv2.resize(im_patch, (int(output_sz[1]), int(output_sz[0])), cv2.INTER_CUBIC)
        if len(im_patch.shape) == 2:
            im_patch = im_patch[:, :, np.newaxis]
        return im_patch 

def random_transform(image, rotation_range, zoom_range, shift_range, random_flip):
    h, w = image.shape[0:2]
    rotation = np.random.uniform(-rotation_range, rotation_range)
    scale = np.random.uniform(1 - zoom_range, 1 + zoom_range)
    tx = np.random.uniform(-shift_range, shift_range) * w
    ty = np.random.uniform(-shift_range, shift_range) * h
    mat = cv2.getRotationMatrix2D((w // 2, h // 2), rotation, scale)
    mat[:, 2] += (tx, ty)
    result = cv2.warpAffine(image, mat, (w, h), borderMode=cv2.BORDER_REPLICATE)
    if np.random.random() < random_flip:
        result = result[:, ::-1]
    return result 

def _augment(self, img, s):
        return cv2.GaussianBlur(img, s, sigmaX=0, sigmaY=0,
                borderType=cv2.BORDER_REPLICATE) 

def padCropImg(img):
    
    H = img.shape[0]
    W = img.shape[1]

    patchRes = 128
    pH = patchRes
    pW = patchRes
    ovlp = int(patchRes * 0.125)

    padH = (int((H - patchRes)/(patchRes - ovlp) + 1) * (patchRes - ovlp) + patchRes) - H
    padW = (int((W - patchRes)/(patchRes - ovlp) + 1) * (patchRes - ovlp) + patchRes) - W

    padImg = cv2.copyMakeBorder(img, 0, padH, 0, padW, cv2.BORDER_REPLICATE)

    ynum = int((padImg.shape[0] - pH)/(pH - ovlp)) + 1
    xnum = int((padImg.shape[1] - pW)/(pW - ovlp)) + 1

    totalPatch = np.zeros((ynum, xnum, patchRes, patchRes, 3), dtype=np.uint8)

    for j in range(0, ynum):
        for i in range(0, xnum):

            x = int(i * (pW - ovlp))
            y = int(j * (pH - ovlp))

            totalPatch[j, i] = padImg[y:int(y + patchRes), x:int(x + patchRes)]

    return totalPatch 

def __init__(self, max_deg, center_range=(0,1),
            interp=cv2.INTER_CUBIC,
            border=cv2.BORDER_REPLICATE):
        """
        :param max_deg: max abs value of the rotation degree
        :param center_range: the location of the rotation center
        """
        super(Rotation, self).__init__()
        self._init(locals()) 

def pad_img_to_fit_bbox(img, x1, x2, y1, y2):
    img = cv.copyMakeBorder(img, - min(0, y1), max(y2 - img.shape[0], 0),
                            -min(0, x1), max(x2 - img.shape[1], 0), cv.BORDER_REPLICATE)
    y2 += -min(0, y1)
    y1 += -min(0, y1)
    x2 += -min(0, x1)
    x1 += -min(0, x1)
    return img, x1, x2, y1, y2 

def wordSegmentation(img, kernelSize=25, sigma=11, theta=7, minArea=0):
	"""Scale space technique for word segmentation proposed by R. Manmatha: http://ciir.cs.umass.edu/pubfiles/mm-27.pdf
	
	Args:
		img: grayscale uint8 image of the text-line to be segmented.
		kernelSize: size of filter kernel, must be an odd integer.
		sigma: standard deviation of Gaussian function used for filter kernel.
		theta: approximated width/height ratio of words, filter function is distorted by this factor.
		minArea: ignore word candidates smaller than specified area.
		
	Returns:
		List of tuples. Each tuple contains the bounding box and the image of the segmented word.
	"""

	# apply filter kernel
	kernel = createKernel(kernelSize, sigma, theta)
	imgFiltered = cv2.filter2D(img, -1, kernel, borderType=cv2.BORDER_REPLICATE).astype(np.uint8)
	(_, imgThres) = cv2.threshold(imgFiltered, 0, 255, cv2.THRESH_BINARY+cv2.THRESH_OTSU)
	imgThres = 255 - imgThres

	# find connected components. OpenCV: return type differs between OpenCV2 and 3
	if cv2.__version__.startswith('3.'):
		(_, components, _) = cv2.findContours(imgThres, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
	else:
		(components, _) = cv2.findContours(imgThres, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)

	# append components to result
	res = []
	for c in components:
		# skip small word candidates
		if cv2.contourArea(c) < minArea:
			continue
		# append bounding box and image of word to result list
		currBox = cv2.boundingRect(c) # returns (x, y, w, h)
		(x, y, w, h) = currBox
		currImg = img[y:y+h, x:x+w]
		res.append((currBox, currImg))

	# return list of words, sorted by x-coordinate
	return sorted(res, key=lambda entry:entry[0][0]) 

def __init__(self, angle_range=(-30.0, 30.0), mode='linear', border='constant', fillval=0, 
                 anchor=None, random_state=np.random):   
        if isinstance(angle_range, numbers.Number):
            angle_range = (-angle_range, angle_range)
        self.angle_range = angle_range

        if isinstance(fillval, numbers.Number):
            fillval = [fillval] * 3

        self.mode = mode
        self.border = BorderTypes.get(border, cv2.BORDER_REPLICATE)
        self.fillval = fillval
        self.anchor = anchor
        self.random = random_state 

def estimate_std(z, method='daub_reflect'):
    # Estimates noise standard deviation assuming additive gaussian noise

    # Check method
    if (method not in NoiseEstMethod.values()) and (method in NoiseEstMethod.keys()):
        method = NoiseEstMethod[method]
    else:
        raise Exception("Invalid noise estimation method.")

    # Check shape
    if len(z.shape) == 2:
        z = z[..., np.newaxis]
    elif len(z.shape) != 3:
        raise Exception("Supports only up to 3D images.")

    # Run on multichannel image
    channels = z.shape[2]
    dev = np.zeros(channels)

    # Iterate over channels
    for ch in range(channels):

        # Daubechies denoising method
        if method == NoiseEstMethod['daub_reflect'] or method == NoiseEstMethod['daub_replicate']:
            daub6kern = np.array([0.03522629188571, 0.08544127388203, -0.13501102001025,
                                  -0.45987750211849, 0.80689150931109, -0.33267055295008],
                                 dtype=np.float32, order='F')

            if method == NoiseEstMethod['daub_reflect']:
                wav_det = cv2.sepFilter2D(z, -1, daub6kern, daub6kern,
                                          borderType=cv2.BORDER_REFLECT_101)
            else:
                wav_det = cv2.sepFilter2D(z, -1, daub6kern, daub6kern,
                                          borderType=cv2.BORDER_REPLICATE)

            dev[ch] = np.median(np.absolute(wav_det)) / 0.6745

    # Return standard deviation
    return dev 

def __call__(self, image, is_mask=False):
        if isinstance(image, torch.Tensor):
            return self.crop_to_output(numpy_to_torch(self(torch_to_numpy(image))))
        else:
            c = (np.expand_dims(np.array(image.shape[:2]),1)-1)/2
            R = np.array([[math.cos(self.angle), math.sin(self.angle)],
                          [-math.sin(self.angle), math.cos(self.angle)]])
            H =np.concatenate([R, c - R @ c], 1)
            return cv.warpAffine(image, H, image.shape[1::-1], borderMode=cv.BORDER_REPLICATE)