Python cv2.invertAffineTransform() 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 rotate_image_with_invrmat(cvmat, rotateAngle):

    assert (cvmat.dtype == np.uint8) , " only support normalize np.uint  in rotate_image_with_invrmat'"

    ##Make sure cvmat is square?
    height, width, channel = cvmat.shape

    center = ( width//2, height//2)
    rotateMatrix = cv2.getRotationMatrix2D(center, rotateAngle, 1.0)

    cos, sin = np.abs(rotateMatrix[0,0]), np.abs(rotateMatrix[0, 1])
    newH = int((height*sin)+(width*cos))
    newW = int((height*cos)+(width*sin))

    rotateMatrix[0,2] += (newW/2) - center[0] #x
    rotateMatrix[1,2] += (newH/2) - center[1] #y

    # rotate image
    outMat = cv2.warpAffine(cvmat, rotateMatrix, (newH, newW), borderValue=(128, 128, 128))

    # generate inv rotate matrix
    invRotateMatrix = cv2.invertAffineTransform(rotateMatrix)

    return (outMat, invRotateMatrix, (width, height)) 

def transformPointsInverse(T,width,height):
    T=cv2.invertAffineTransform(T)
    return transformPointsForward(T,width,height) 

def affine_skew(tilt, phi, img, mask=None):
    '''
    affine_skew(tilt, phi, img, mask=None) -> skew_img, skew_mask, Ai

    Ai - is an affine transform matrix from skew_img to img
    '''
    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 Alignment_2(img,std_landmark,landmark):
    def Transformation(std_landmark,landmark):
        std_landmark = np.matrix(std_landmark).astype(np.float64)
        landmark = np.matrix(landmark).astype(np.float64)

        c1 = np.mean(std_landmark, axis=0)
        c2 = np.mean(landmark, axis=0)
        std_landmark -= c1
        landmark -= c2

        s1 = np.std(std_landmark)
        s2 = np.std(landmark)
        std_landmark /= s1
        landmark /= s2 

        U, S, Vt = np.linalg.svd(std_landmark.T * landmark)
        R = (U * Vt).T

        return np.vstack([np.hstack(((s2 / s1) * R, c2.T - (s2 / s1) * R * c1.T)),np.matrix([0., 0., 1.])])

    Trans_Matrix = Transformation(std_landmark,landmark) # Shape: 3 * 3
    Trans_Matrix = Trans_Matrix[:2]
    Trans_Matrix = cv2.invertAffineTransform(Trans_Matrix)
    new_img = cv2.warpAffine(img,Trans_Matrix,(img.shape[1],img.shape[0]))

    Trans_Matrix = np.array(Trans_Matrix)
    new_landmark = []
    for i in range(landmark.shape[0]):
        pts = []    
        pts.append(Trans_Matrix[0,0]*landmark[i,0]+Trans_Matrix[0,1]*landmark[i,1]+Trans_Matrix[0,2])
        pts.append(Trans_Matrix[1,0]*landmark[i,0]+Trans_Matrix[1,1]*landmark[i,1]+Trans_Matrix[1,2])
        new_landmark.append(pts)

    new_landmark = np.array(new_landmark)

    return new_img, new_landmark

#---------------------------------#
#   图片预处理
#   高斯归一化
#---------------------------------# 

def extract_box(img, box, padding_factor = 0.2):
    '''
    we can search for whatever we want in the rotated bordered image, 
    
    Any point found can be translated back to the original image by:
    1. adding the origins of the bordered area,
    2. rotating the point using the inverse rotation matrix    
    
    '''
    
    if box.angle != 0:
        
        b_w = max(img.shape)*2
        b_h = b_w
        dx_center = b_w / 2 - box.center[0]
        dy_center = b_h / 2 - box.center[1]
        new_img = np.zeros((b_w, b_h, 3), dtype = img.dtype)
        new_img[dy_center:(dy_center + img.shape[0]), dx_center:(dx_center + img.shape[1]), :] = img
        
        box_in_big_image = box.points + np.c_[np.ones((4,1)) * dx_center, np.ones((4,1)) * dy_center]

        rot_mat = cv2.getRotationMatrix2D((b_w/2, b_h/2), box.angle, scale = 1.0)
        inv_rot_mat = cv2.invertAffineTransform(rot_mat)
        rot_image = cv2.warpAffine(new_img, rot_mat, (new_img.shape[1],new_img.shape[0]), flags=cv2.INTER_CUBIC)
        box_UL_in_rotated = (rot_mat * np.matrix([box_in_big_image[0,0], box_in_big_image[0,1], 1]).transpose()).transpose().tolist()[0] 
        box_coords_in_rotated = np.matrix(np.c_[box_in_big_image, np.ones((4,1))]) * rot_mat.T
        box_coords_in_rotated = box_coords_in_rotated[0,:].tolist()[0] + [box.dx, box.dy]
    else:
        rot_mat = cv2.getRotationMatrix2D(box.center, box.angle, scale = 1.0)
        inv_rot_mat = cv2.invertAffineTransform(rot_mat)
        # for efficiency
        rot_image = img.copy()
        box_UL_in_rotated = (rot_mat * np.matrix([box.points[0,0], box.points[0,1], 1]).transpose()).transpose().tolist()[0] 
        box_coords_in_rotated = box_UL_in_rotated + [box.dx, box.dy]
    
    img_with_border, Dx, Dy = extract_rect(rot_image, box_coords_in_rotated, padding_factor)
    box_coords_in_bordered = [Dx, Dy] + [box.dx, box.dy]
    border_UL_in_rotated = [box_UL_in_rotated[0]-Dx, box_UL_in_rotated[1]-Dy]
    
    return img_with_border, box_coords_in_bordered, border_UL_in_rotated, inv_rot_mat 

def original_roi(self):
        """ :class: `numpy.ndarray`: The original region of interest of the mask in the
        source frame. """
        points = np.array([[0, 0],
                           [0, self.stored_size - 1],
                           [self.stored_size - 1, self.stored_size - 1],
                           [self.stored_size - 1, 0]], np.int32).reshape((-1, 1, 2))
        matrix = cv2.invertAffineTransform(self._affine_matrix)
        roi = cv2.transform(points, matrix).reshape((4, 2))
        logger.trace("Returning: %s", roi)
        return roi 

def get_original_roi(self, mat, size, padding=0):
        """ Return the square aligned box location on the original image """
        logger.trace("matrix: %s, size: %s. padding: %s", mat, size, padding)
        matrix = self.transform_matrix(mat, size, padding)
        points = np.array([[0, 0], [0, size - 1], [size - 1, size - 1], [size - 1, 0]], np.int32)
        points = points.reshape((-1, 1, 2))
        matrix = cv2.invertAffineTransform(matrix)
        logger.trace("Returning: (points: %s, matrix: %s", points, matrix)
        return cv2.transform(points, matrix) 

def _rotate_face(face, rotation_matrix):
        """ Rotates the detection bounding box around the given rotation matrix.

        Parameters
        ----------
        face: :class:`DetectedFace`
            A :class:`DetectedFace` containing the `x`, `w`, `y`, `h` detection bounding box
            points.
        rotation_matrix: numpy.ndarray
            The rotation matrix to rotate the given object by.

        Returns
        -------
        :class:`DetectedFace`
            The same class with the detection bounding box points rotated by the given matrix.
        """
        logger.trace("Rotating face: (face: %s, rotation_matrix: %s)", face, rotation_matrix)
        bounding_box = [[face.left, face.top],
                        [face.right, face.top],
                        [face.right, face.bottom],
                        [face.left, face.bottom]]
        rotation_matrix = cv2.invertAffineTransform(rotation_matrix)

        points = np.array(bounding_box, "int32")
        points = np.expand_dims(points, axis=0)
        transformed = cv2.transform(points, rotation_matrix).astype("int32")
        rotated = transformed.squeeze()

        # Bounding box should follow x, y planes, so get min/max for non-90 degree rotations
        pt_x = min([pnt[0] for pnt in rotated])
        pt_y = min([pnt[1] for pnt in rotated])
        pt_x1 = max([pnt[0] for pnt in rotated])
        pt_y1 = max([pnt[1] for pnt in rotated])
        width = pt_x1 - pt_x
        height = pt_y1 - pt_y

        face.x = int(pt_x)
        face.y = int(pt_y)
        face.w = int(width)
        face.h = int(height)
        return face 

def transform_points(points, mat, invert=False):
    if invert:
        mat = cv2.invertAffineTransform (mat)
    points = np.expand_dims(points, axis=1)
    points = cv2.transform(points, mat, points.shape)
    points = np.squeeze(points)
    return points 

def affine_skew(tilt, phi, img, mask=None):
    '''
    affine_skew(tilt, phi, img, mask=None) -> skew_img, skew_mask, Ai

    Ai - is an affine transform matrix from skew_img to img
    '''
    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 affine_skew(tilt, phi, img, mask=None):
    '''
    affine_skew(tilt, phi, img, mask=None) -> skew_img, skew_mask, Ai

    Ai - is an affine transform matrix from skew_img to img
    '''
    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 face_swap(orig_image, down_scale):
    # extract face from original
    facelist = extract_faces(orig_image, 256)
    result_image = orig_image

    # iterate through all detected faces
    for (face, resized_image) in facelist:
        range_ = numpy.linspace( 128-80, 128+80, 5 )
        mapx = numpy.broadcast_to( range_, (5,5) )
        mapy = mapx.T

        # warp image like in the training
        mapx = mapx + numpy.random.normal( size=(5,5), scale=5 )
        mapy = mapy + numpy.random.normal( size=(5,5), scale=5 )

        src_points = numpy.stack( [ mapx.ravel(), mapy.ravel() ], axis=-1 )
        dst_points = numpy.mgrid[0:65:16,0:65:16].T.reshape(-1,2)
        mat = umeyama( src_points, dst_points, True )[0:2]

        warped_resized_image = cv2.warpAffine( resized_image, mat, (64,64) ) / 255.0

        test_images = numpy.empty( ( 1, ) + warped_resized_image.shape )
        test_images[0] = warped_resized_image

        # predict faceswap using encoder A
        figure = autoencoder_A.predict(test_images)

        new_face = numpy.clip(numpy.squeeze(figure[0]) * 255.0, 0, 255).astype('uint8')
        mat_inv = umeyama( dst_points, src_points, True )[0:2]

        # insert face into extracted face
        dest_face = blend_warp(new_face, resized_image, mat_inv)

        # create an inverse affine transform matrix to insert extracted face again
        mat = get_align_mat(face)
        mat = mat * (256 - 2 * 48)
        mat[:,2] += 48    
        mat_inv = cv2.invertAffineTransform(mat)
        # insert new face into original image
        result_image = blend_warp(dest_face, result_image, mat_inv)

    # return resulting image after downscale
    return cv2.resize(result_image, (result_image.shape[1] // down_scale, result_image.shape[0] // down_scale))