@property
def width(self):
    """Return the width (int) of the image."""
    w = self.shape[1]
    return w

@property
def height(self):
    """Return the height (int) of the image."""
    h = self.shape[0]
    return h

@property
def channel(self):
    """Return the number of channels (int) of the image."""
    if len(self.shape) > 2:
        _, _, c = self.shape
    else:
        c = 0
    return c

@property
def center(self):
    """Return the position tuple(x, y) of the center of the image."""
    w, h = self.size()
    return (w // 2, h // 2)

def size(self):
    """Return the size of the image as tuple(width, height)."""
    h, w = self.shape[:2]
    return (w, h)

def euclidian_distance(self, other):
    """Return the euclidian distance with the image.

        :param other: (sppasImage) an image with the same shape

        """
    w, h = self.size()
    d = numpy.linalg.norm(self - other, axis=1)
    return sum(sum(d)) / (w * h * 3)

def get_proportional_size(self, width=0, height=0):
    """Return the size of the image or a proportional size.

        :param width: (int) Force the image to the width
        :param height: (int) Force the image to the height
        :return: tuple(int, int) Width and height

        """
    if len(self) == 0:
        return (0, 0)
    width = sppasCoords.to_dtype(width)
    height = sppasCoords.to_dtype(height)
    if width < 0:
        raise NegativeValueError(width)
    if height < 0:
        raise NegativeValueError(height)
    h, w = self.shape[:2]
    if width + height == 0:
        return (w, h)
    prop_width = prop_height = 0
    propw = proph = 1.0
    if width != 0:
        prop_width = width
        propw = float(width) / float(w)
    if height != 0:
        prop_height = height
        proph = float(height) / float(h)
    if width == 0:
        prop_width = int(float(w) * proph)
    if height == 0:
        prop_height = int(float(h) * propw)
    return (prop_width, prop_height)

def surround_coord(self, coord, color, thickness, text=''):
    """Add a square surrounding the given coordinates.

        :param coord: (sppasCoords) Area to surround
        :param color: (int, int, int) Rectangle color or brightness (if grayscale image).
        :param thickness: (int) Thickness of lines that make up the rectangle.
            Negative values, like CV_FILLED , mean that the function has to
            draw a filled rectangle.
        :param text: (str) Add text

        """
    coord = sppasCoords.to_coords(coord)
    cv2.rectangle(self, (coord.x, coord.y), (coord.x + coord.w, coord.y + coord.h), color, thickness)
    if len(text) > 0:
        h, w = self.shape[:2]
        font_scale = float(w * h) / (1920.0 * 1080.0)
        th = abs(thickness // 2)
        text_size = cv2.getTextSize(text, fontFace=cv2.FONT_HERSHEY_SIMPLEX, fontScale=font_scale * 2, thickness=th)
        if thickness < 0:
            r, g, b = color
            r = (r + 128) % 255
            g = (g + 128) % 255
            b = (b + 128) % 255
            color = (r, g, b)
        cv2.putText(self, text, (coord.x + 3 * th, coord.y + 3 * th + text_size[1]), cv2.FONT_HERSHEY_SIMPLEX, font_scale, color, th)

def put_text(self, coord, color, thickness, text):
    """Put a text at the given coords.

        :param coord: (sppasCoords) Area to put the text
        :param color: (int, int, int) Rectangle color or brightness (if grayscale image).
        :param thickness: (int) Thickness of lines that make up the rectangle.
            Negative values, like CV_FILLED , mean that the function has to
            draw a filled rectangle.
        :param text: (str) Add text

        """
    w, _ = self.size()
    coord = sppasCoords.to_coords(coord)
    font_scale = max(1.0, float(w) / 1024.0)
    if thickness < 0:
        r, g, b = color
        r = (r + 128) % 255
        g = (g + 128) % 255
        b = (b + 128) % 255
        color = (r, g, b)
    cv2.putText(self, text, (coord.x, coord.y), cv2.FONT_HERSHEY_DUPLEX, font_scale, color, thickness)

def surround_point(self, point, color, thickness, radius=None):
    """Add a circle surrounding the given point.

        :param point: (sppasCoords, list, tuple) (x,y) values to surround
        :param color: (int, int, int) Rectangle color or brightness (if grayscale image).
        :param thickness: (int) Thickness of lines that make up the rectangle.
            Negative values, like CV_FILLED , mean that the function has to
            draw a filled rectangle.
        :param radius: (int) Radius of the circle

        """
    if isinstance(point, sppasCoords) is False:
        if isinstance(point, (tuple, list)) and len(point) >= 2:
            try:
                point = sppasCoords(point[0], point[1])
            except Exception:
                pass
    if isinstance(point, sppasCoords) is False:
        sppasTypeError(point, 'sppasCoords, tuple, list')
    thickness = max(2, thickness)
    if radius is None:
        radius = thickness // 2
    x = point.x - radius
    y = point.y - radius
    cv2.circle(self, (x, y), radius, color, thickness)

def blank_image(self, w=0, h=0, white=False, alpha=None, dtype=numpy.uint8):
    """Create and return an image with black pixels only.

        :param w: (int) Image width. 0 means to use the current image width.
        :param h: (int) Image height. 0 means to use the current height.
        :param white: (bool) Return a white image instead of a black one.
        :param alpha: (int | None) Add alpha channel with the given int value
        :param dtype: (numpy.dtype) Image data type
        :return: (sppasImage) Fully black BGR image

        """
    if w < 0:
        raise NegativeValueError(w)
    if h < 0:
        raise NegativeValueError(h)
    if w == 0:
        w = self.width
    if h == 0:
        h = self.height
    t = (h, w, 3)
    nparray = numpy.zeros(t, dtype=dtype)
    if white is True:
        nparray[:, :, (0, 1, 2)] = 255
    img = sppasImage(input_array=nparray)
    if alpha is not None:
        return img.ialpha(alpha)
    return img

def ired(self, value=0):
    """Return a copy of the image in red-color.

        :param value: (int) Fixed red value ranging (0, 255)
        :return: (sppasImage)

        """
    value = int(value)
    value = value % 255
    img = self.copy()
    img[:, :, (0, 1)] = value
    return sppasImage(input_array=img)

def igreen(self, value=0):
    """Return a copy of the image in green-color.

        :param value: (int) Fixed green value ranging (0, 255)
        :return: (sppasImage)

        """
    value = int(value)
    value = value % 255
    img_green = self.copy()
    img_green[:, :, (0, 2)] = value
    return sppasImage(input_array=img_green)

def iblue(self, value=0):
    """Return a copy of the image in blue-color.

        :param value: (int) Fixed blue value ranging (0, 255)
        :return: (sppasImage)

        """
    value = int(value)
    value = value % 255
    img = self.copy()
    img[:, :, (1, 2)] = value
    return sppasImage(input_array=img)

def ialpha(self, value=0, direction=0):
    """Return a copy of the image in RGBA colors.

        Do nothing if no channel defined.

        :param value: (int) Alpha value for transparency (0-255)
        :param direction: (int) 0 means to assign the value to each pixel, but
        -1 means to only assign the value to pixels if the existing transparency
        is higher than value (lowers are un-changed) and +1 means to assign
        alpha value if the existing one is lower (higher values are unchanged).
        :return: (sppasImage)

        """
    if self.channel == 3:
        imga = sppasImage(input_array=cv2.cvtColor(self, cv2.COLOR_RGB2RGBA))
    else:
        imga = self.copy()
    if imga.channel == 4:
        value = int(value)
        value = value % 255
        if direction == 0:
            imga[:, :, 3] = value
        elif direction < 0:
            numpy.clip(imga[:, :, 3], 0, value, out=imga[:, :, 3])
        else:
            numpy.clip(imga[:, :, 3], value, 255, out=imga[:, :, 3])
    return imga

def ibgr(self, bgr):
    """Return a copy of the image with given in BGR/BGRA color values.

        :param bgr: (tuple) A tuple(b, g, r) or tuple(b, g, r, a) color.
        :return: (sppasImage)

        """
    if len(bgr) < 3:
        raise ValueError('Expected a (b, g, r) or (b, g, r, a) color. Got {} instead.'.format(bgr))
    img = self.copy()
    value = int(bgr[0])
    value = value % 255
    img[:, :, 0] = value
    value = int(bgr[1])
    value = value % 255
    img[:, :, 1] = value
    value = int(bgr[2])
    value = value % 255
    img[:, :, 2] = value
    if len(bgr) == 4:
        return img.ialpha(bgr[3])
    return img

def ibgra_to_bgr(self):
    """Return a copy of the image without the alpha channel.

        :return: (sppasImage)

        """
    return sppasImage(input_array=cv2.cvtColor(self, cv2.COLOR_BGRA2BGR))

def igray(self):
    """Return a copy of the image in grayscale.

        :return: (sppasImage)

        """
    if self.channel == 4:
        avg = numpy.average(self, weights=[0.114, 0.587, 0.2989, 1], axis=2)
    elif self.channel == 3:
        avg = numpy.average(self, weights=[0.114, 0.587, 0.2989], axis=2)
    else:
        return self.copy()
    gray = self.copy()
    gray[:, :, 0] = avg
    gray[:, :, 1] = avg
    gray[:, :, 2] = avg
    return sppasImage(input_array=gray)

def inegative(self):
    """Return a copy of the image in negative/positive colors.

        :return: (sppasImage)

        """
    img = 255 - self
    return sppasImage(input_array=img)

def ireduction(self, value=128):
    """Return a copy of the image with a color-reduction applied.

        :param value: (int) Reduction value in range(0, 255)
        :return: (sppasImage)

        """
    value = int(value)
    if value < 0:
        return self.copy()
    coeff = value % 255
    img = self // coeff * coeff
    return sppasImage(input_array=img)

def igamma(self, coeff=1.0):
    """Return a copy of the image with lightness changed.

        :param coeff: (float) Set a value in range (0, 1) to increase
            lightness, or a value > 1 to increase darkness.
        :return: (sppasImage)

        """
    if coeff < 0.0:
        coeff = 0.0
    img = 255.0 * (self / 255.0) ** coeff
    return sppasImage(input_array=img)

def ito_rgb(self):
    """Return a copy of the image representing RGB colors.

        Remove alpha channel if any.

        :raises: TypeError
        :return: (sppasImage) an RGB image

        """
    img = self.copy()
    if self.channel > 2:
        return img[:, :, [2, 1, 0]]
    raise sppasTypeError('image', 'BGR/BGRA sppasImage')

def ishadow(self, x=5, y=20):
    """Return a copy of image with a shadow added.

        :param x: (int) Shadow width
        :param y: (int) Shadow height
        :return: (sppasImage)

        """
    tmp = self.igray()
    tmp = tmp.ishift(x, y)
    return tmp.ioverlay(self, (0, 0))

def icrop(self, coord):
    """Return a trimmed part of the image at given coordinates.

        :param coord: (sppasCoords) crop to these x, y, w, h values.
        :return: (sppasImage)

        """
    coord = sppasCoords.to_coords(coord)
    x1 = coord.x
    x2 = coord.x + coord.w
    y1 = coord.y
    y2 = coord.y + coord.h
    cropped = self[y1:y2, x1:x2]
    return sppasImage(input_array=cropped)

def itrim(self, coord):
    """Return a cropped part of the image.

        :param coord: (tuple or sppasCoords)
        :return: (sppasImage)

        """
    return self.icrop(coord)

def iresize(self, width=0, height=0):
    """Return a copy of the image with the specified width and height.

        :param width: (int) The width to resize to (0=proportional to height)
        :param height: (int) The height to resize to (0=proportional to width)
        :return: (sppasImage)

        """
    if width == 0 and height == 0:
        return self.copy()
    prop_width, prop_height = self.get_proportional_size(width, height)
    if prop_width + prop_height == 0:
        return self.copy()
    dif = self.height if self.height > self.width else self.width
    interpol = cv2.INTER_AREA if dif > (width + height) // 2 else cv2.INTER_CUBIC
    image = cv2.resize(self, (prop_width, prop_height), interpolation=interpol)
    return sppasImage(input_array=image)

def izoom(self, width, height):
    """Return a zoomed copy of the image.

        Resize and crop the image to zoom it to the given size.
        Keep the original aspect ratio of the image, but crop if necessary.

        :param width: (int) The width to resize to
        :param height: (int) The height to resize to
        :return: (sppasImage)

        """
    aspect_ratio = int(100.0 * float(self.width) / float(self.height)) / 100.0
    res_aspect_ratio = int(100.0 * float(width) / float(height)) / 100.0
    if aspect_ratio > res_aspect_ratio:
        img_w = int(aspect_ratio * float(height))
        img_h = height
        img = self.iresize(img_w, img_h)
        x1 = int(float(img_w - width) / 2.0)
        x2 = x1 + width
        img = img[:, x1:x2, :]
    elif aspect_ratio < res_aspect_ratio:
        img_w = width
        img_h = int(float(width) / aspect_ratio)
        img = self.iresize(img_w, img_h)
        y1 = int(float(img_h - height) / 2.0)
        y2 = y1 + height
        img = img[y1:y2, :, :]
    else:
        img = self.iresize(width, height)
    return sppasImage(input_array=img)

def icenter(self, width, height):
    """Return a copy of the image centered in an image of given size.

        Center the image into a blank image of the given size.
        Keep the original aspect ratio of the image, crop if necessary or
        add a black border all around.

        :param width: (int) The width to resize to
        :param height: (int) The height to resize to
        :return: (sppasImage)

        """
    coord = sppasCoords(0, 0, width, height)
    if self.width > width:
        coord.x = (self.width - width) // 2
    if self.height > height:
        coord.y = (self.height - height) // 2
    img = self.icrop(coord)
    if self.channel == 4:
        mask = sppasImage(0).blank_image(width, height, white=False, alpha=0)
    else:
        mask = sppasImage(0).blank_image(width, height)
    x_pos = (width - img.width) // 2
    y_pos = (height - img.height) // 2
    mask[y_pos:y_pos + img.height, x_pos:x_pos + img.width, :] = img[:img.height, :img.width, :]
    return sppasImage(input_array=mask)

def iextend(self, width, height):
    """Return an extended copy of the image of given size.

        Scale the image to match the given size, keeping aspect ratio.
        Keep the original aspect ratio of the image, add a black border.

        :param width: (int) The width to resize to
        :param height: (int) The height to resize to
        :return: (sppasImage)

        """
    aspect_ratio = int(100.0 * float(self.width) / float(self.height)) / 100.0
    res_aspect_ratio = int(100.0 * float(width) / float(height)) / 100.0
    if aspect_ratio == res_aspect_ratio:
        return self.iresize(width, height)
    elif aspect_ratio > res_aspect_ratio:
        coeff = float(width) / float(self.width)
        img = self.iresize(width, int(coeff * float(self.height)))
    else:
        coeff = float(height) / float(self.height)
        img = self.iresize(int(coeff * float(self.width)), height)
    return img.icenter(width, height)

def irotate(self, angle, center=None, scale=1.0, redimension=True):
    """Return a rotated copy of the image with the given angle.

        This method is part of imutils under the terms of the MIT License (MIT)
        Copyright (c) 2015-2016 Adrian Rosebrock, http://www.pyimagesearch.com
        See here for details:
        https://www.pyimagesearch.com/2017/01/02/rotate-images-correctly-with-opencv-and-python/

        :param angle: (float) Rotation angle in degrees.
        :param center: (int) Center of the rotation in the source image.
        :param scale: (float) Isotropic scale factor.
        :param redimension: (bool) Scale the image to fit in the previous square
        :return: (sppasImage)

        """
    h, w = self.shape[:2]
    if center is None:
        center = (w // 2, h // 2)
    matrix = cv2.getRotationMatrix2D(center, -angle, scale)
    if redimension is True:
        cos = numpy.abs(matrix[0, 0])
        sin = numpy.abs(matrix[0, 1])
        new_width = int(h * sin + w * cos)
        new_height = int(h * cos + w * sin)
        matrix[0, 2] += new_width // 2 - center[0]
        matrix[1, 2] += new_height // 2 - center[1]
    else:
        new_width = w
        new_height = h
    rotated = cv2.warpAffine(self, matrix, (new_width, new_height))
    return sppasImage(input_array=rotated)

def ishift(self, x, y):
    """Return a shifted copy of the image at given coordinates.

        Shift the content at left/right and/or top/bottom.

        :param x: (int)
        :param y: (int)
        :return: (sppasImage)

        """
    h, w = self.shape[:2]
    matrix = numpy.float32([[1, 0, x], [0, 1, y]])
    shifted = cv2.warpAffine(self, matrix, (w, h))
    return sppasImage(input_array=shifted)

def iflip(self, flip_code=-1):
    """Return a flipped copy of the image.

        flip_code = 0: flip vertically
        flip_code > 0: flip horizontally
        flip_code < 0: flip vertically and horizontally

        :param flip_code: (int) Indicate the way to flip the image
        :return: (sppasImage)

        """
    if flip_code == 0:
        img = numpy.flipud(self)
    elif flip_code > 0:
        img = numpy.fliplr(self)
    else:
        img = numpy.flip(self, (0, 1))
    return sppasImage(input_array=img)

def imask(self, other):
    """Return a copy of the image masked with the other given image.

        :param other: (sppasImage) Image to mask (black areas)
        :return: (sppasImage)

        """
    w, h = self.size()
    other = other.iresize(w, h)
    dst = self * other / 255
    dst.astype(numpy.uint8)
    return sppasImage(input_array=dst)

def iblur(self, value=51, method='gaussian'):
    """Return a copy of the image with smoothed borders.

        :param value: (int) Kernel value, from 0 to 51.
        :param method: (str) One of: None, "gaussian" or "median"
        :return: (sppasImage)

        """
    if method is None:
        mask_blur = cv2.blur(self, (value, value))
    elif method.lower() == 'median':
        mask_blur = cv2.medianBlur(self, value)
    else:
        mask_blur = cv2.GaussianBlur(self, (value, value), cv2.BORDER_DEFAULT)
    return sppasImage(input_array=mask_blur)

def icontours(self, threshold=128, color=(0, 255, 0)):
    """Return a blank image with the contours of the image in color.

        :param threshold: (int) value which is used to classify the pixel values (0-255)
        :param color: (tuple) BGR values of the color to draw the contours
        :return: (sppasImage)

        """
    if self.channel < 2:
        logging.error('Invalid image to draw a contour.')
        return self
    img_grey = cv2.cvtColor(self, cv2.COLOR_BGR2GRAY)
    _, thresh_img = cv2.threshold(img_grey, threshold, 255, cv2.THRESH_BINARY)
    contours, _ = cv2.findContours(thresh_img, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
    img_contours = numpy.zeros(self.shape)
    cv2.drawContours(img_contours, contours, -1, color, 3)
    return sppasImage(input_array=img_contours)

def ipaste(self, other, coord):
    """Return a copy of the image with other paste on it at given coords.

        Replace the current image with the given one at given coords.

        :param other: (sppasImage) Image to paste
        :param coord: (sppasCoords) Position and optionally size to paste
        :return: (sppasImage)

        """
    img = self.copy()
    other = sppasImage(input_array=other)
    coord = sppasCoords.to_coords(coord)
    if other.channel != self.channel:
        if other.channel == 3 and self.channel == 4:
            other = other.ialpha(0)
        if other.channel == 4 and self.channel == 3:
            img = self.ialpha(0)
    if coord.w > 0 and coord.h > 0:
        paste_img = other.iresize(coord.w, coord.h)
    else:
        paste_img = other
        w, h = other.size()
        coord.w = w
        coord.h = h
    w, h = self.size()
    to_crop = False
    if coord.x + coord.w > w:
        new_w = w - coord.x
        coord.w = new_w
        to_crop = True
    if coord.y + coord.h > h:
        new_h = h - coord.y
        to_crop = True
        coord.h = new_h
    if to_crop is True:
        paste_img = paste_img.icrop((0, 0, coord.w, coord.h))
    x1 = coord.x
    x2 = coord.x + coord.w
    y1 = coord.y
    y2 = coord.y + coord.h
    img[y1:y2, x1:x2] = paste_img
    return img

def iblend(self, other, coord=None, weight1=0.5, weight2=0.5):
    """Return a copy of the image with the other image added or blended.

        :param other: (sppasImage) Image to blend with
        :param coord: (sppasCoord) Blend only the given area of self with other
        :param weight1: (float) coeff on the image
        :param weight2: (float) coeff on the other image
        :return: (sppasImage)

        """
    w, h = self.size()
    img = self.copy()
    if coord is None:
        other = other.iresize(w, h)
    else:
        blank = sppasImage(0).blank_image(w, h, white=False, alpha=0)
        if other.channel == 3:
            other = other.ialpha(254)
        other = blank.ipaste(other, coord)
        if self.channel == 3:
            img = self.ialpha(254)
    blended = cv2.addWeighted(img, weight1, other, weight2, 0)
    return sppasImage(input_array=blended)

def ioverlay(self, other, coord):
    """Return a copy of the image with the other image added.

        :param other: (sppasImage) Image to blend with
        :param coord: (sppasCoord) Overlay in the given area of self with other
        :return: (sppasImage)

        """
    back_image = self.copy()
    w, h = self.size()
    over_image = sppasImage(input_array=other)
    coord = sppasCoords.to_coords(coord)
    if over_image.channel == 3:
        over_image = over_image.ialpha(254)
    if back_image.channel == 3:
        back_image = self.ialpha(254)
    if coord.w > 0 and coord.h > 0:
        over_image = over_image.iresize(coord.w, coord.h)
    cols, rows = over_image.shape[:2]
    x = coord.x
    y = coord.y
    if x + rows > w:
        rows = w - x
    if y + cols > h:
        cols = h - y
    tmp = sppasImage(0).blank_image(w, h, white=False, alpha=254)
    over_image = tmp.ipaste(over_image, (x, y))
    alpha_background = back_image[:, :, 3] / 255.0
    alpha_foreground = over_image[:, :, 3] / 255.0
    roi_over = back_image.copy()
    for color in range(0, 3):
        roi_over[:, :, color] = alpha_foreground * over_image[:, :, color] + alpha_background * roi_over[:, :, color] * (1 - alpha_foreground)
    roi_over[:, :, 3] = (1 - (1 - alpha_foreground) * (1 - alpha_background)) * 255
    if rows > 0 and cols > 0:
        black = sppasImage(0).blank_image(rows, cols, white=False, alpha=254)
        back_image = back_image.ipaste(black, (x, y))
    combined = cv2.add(back_image, roi_over)
    return sppasImage(input_array=combined)

def isurround(self, coords, color=(50, 100, 200), thickness=2, score=False):
    """Return a new image with a square surrounding all the given coords.

        :param coords: (List of sppasCoords) Areas to surround
        :param color: (int, int, int) Rectangle color
        :param thickness: (int) Thickness of lines that make up the rectangle.
            Negative values, like CV_FILLED , mean that the function has to
            draw a filled rectangle.
        :param score: (bool) Add the confidence score of the coords
        :return: (sppasImage)

        """
    img = self.copy()
    for c in coords:
        c = sppasCoords.to_coords(c)
        if c.w > 0 and c.h > 0:
            text = ''
            if score is True and c.get_confidence() > 0.0:
                text = '{:.3f}'.format(c.get_confidence())
            img.surround_coord(c, color, thickness, text)
        else:
            img.surround_point(c, color, thickness)
    return img

def icartoon(self, colorize=True):
    """Applied a filter on the image to create a cartoon effect.

        :param colorize: (bool) If we want our image in only black and
            white (False) or with colors (True).
        :return: (sppasImage) The image with the filter applied on it

        """
    alpha_image = None
    if self.channel == 4:
        alpha_image = self[:, :, 3] / 255.0
        img = self.ibgra_to_bgr()
    else:
        img = self.copy()
    img_blur = cv2.medianBlur(img, 5)
    img_bf = cv2.bilateralFilter(img_blur, d=3, sigmaColor=50, sigmaSpace=50)
    img_lp_al = cv2.Laplacian(img_bf, cv2.CV_8U, ksize=5)
    img_lp_al_grey = cv2.cvtColor(img_lp_al, cv2.COLOR_BGR2GRAY)
    blur_al = cv2.GaussianBlur(img_lp_al_grey, (5, 5), 0)
    _, tresh_al = cv2.threshold(blur_al, 245, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
    inverted_bilateral = cv2.subtract(255, tresh_al)
    if colorize is True:
        img_color = cv2.bilateralFilter(img, d=8, sigmaColor=250, sigmaSpace=250)
        img_color = sppasImage(input_array=cv2.bitwise_and(img_color, img_color, mask=inverted_bilateral))
        if alpha_image is not None:
            img_color = img_color.ialpha(254)
            img_color[:, :, 3] = alpha_image * 255
        return img_color
    if alpha_image is not None:
        img = sppasImage(input_array=inverted_bilateral).ialpha(254)
        img[:, :, 3] = alpha_image * 255
        return img
    else:
        return sppasImage(input_array=inverted_bilateral)

def iquantization_color(self, nb_down_samp=2):
    """Smooth colors, down-sample and Up-sample the original image colors.

        :param nb_down_samp: (int) number of downscaling steps

        :return: (sppasImage) The image with the filter applied on it

        """
    img_color = self.copy()
    for _ in range(nb_down_samp):
        img_color = cv2.pyrDown(img_color)
    for _ in range(50):
        img_color = cv2.bilateralFilter(img_color, 9, 9, 7)
    for _ in range(nb_down_samp):
        img_color = cv2.pyrUp(img_color)
    return sppasImage(input_array=img_color)

def iinvert(self):
    """Applied an invert filter on the image.

        :return: (sppasImage) The image with the filter applied on it

        """
    "\n        alpha_image = None\n\n        # image doesn't have an alpha channel\n        if self.channel == 3:\n            img = self.copy()\n        else:\n            # normalize alpha channels from 0-255 to 0-1\n            alpha_image = self[:, :, 3] / 255.0\n            # get the image without the alpha channel\n            img = self.ibgra_to_bgr()\n\n        # apply the invert filter\n        invert_array = cv2.bitwise_not(img)\n\n        if alpha_image is not None:\n            # create alpha channel and restore original image alpha\n            img = sppasImage(input_array=invert_array).ialpha(254)\n            img[:, :, 3] = alpha_image * 255\n\n        return img\n        "
    invert_array = cv2.bitwise_not(self.copy())
    return sppasImage(input_array=invert_array)

def ioverlay_color(self, color_overlay, intensity=0.2):
    """Applied an overlay with a specified color on the image.

        :param color_overlay: (tuple[int, int, int]) The color of the overlay on the rgb format
        :param intensity: (float) The intensity

        :return: (sppasImage) The image with the filter applied on it

        """
    image = cv2.cvtColor(self.copy(), cv2.COLOR_BGR2BGRA)
    image_height, image_width, _ = image.shape
    color_bgra = color_overlay[::-1] + (1,)
    overlay = numpy.full((image_height, image_width, 4), color_bgra, dtype='uint8')
    cv2.addWeighted(overlay, intensity, image, 1.0, 0, image)
    image = cv2.cvtColor(image, cv2.COLOR_BGRA2BGR)
    return sppasImage(input_array=image)

def iuint8(self):
    """Return a copy of the image with dtype=numpy.uint8."""
    img = self / self.max()
    img = 255 * img
    return img.astype(numpy.uint8)

def write(self, filename):
    """Write the image on disk.

        :param filename: (str) Name of the image file
        :raises: (ImageWriteError)

        """
    try:
        cv2.imwrite(filename, self)
    except cv2.error as e:
        logging.error('Error when writing file {}: {}'.format(filename, str(e)))
        raise ImageWriteError(filename)

@staticmethod
def _to_image(entry):
    """Return a sppasImage from the given entry.

        :param entry: (numpy.ndarray or filename)
        :return: (sppasImage)

        """
    if isinstance(entry, sppasImage) is True:
        return entry
    if isinstance(entry, numpy.ndarray):
        return sppasImage(input_array=entry)
    return sppasImage(filename=entry)

def __eq__(self, other):
    """Allows to write img1 == img2."""
    if len(self) != len(other):
        return False
    for l1, l2 in zip(self, other):
        if len(l1) != len(l2):
            return False
        for c1, c2 in zip(l1, l2):
            if len(c1) != len(c2):
                return False
            r1, g1, b1 = c1
            r2, g2, b2 = c2
            if r1 != r2 or g1 != g2 or b1 != b2:
                return False
    return True

def __ne__(self, other):
    """Allows to write img1 != img2."""
    return not self.__eq__(other)