pypocquant.lib.processing

Module Contents

Functions

phase_only_correlation(in1: np.ndarray, in2: np.ndarray) → np.ndarray

Calculate phase-only correlation of two numpy arrays.

find_position_in_image_using_phase_corr(in1: np.ndarray, in2: np.ndarray) → tuple

Uses phase-only correlation to find the coordinates in in2 where in1 can be found.

find_position_in_image_using_norm_xcorr(in1: np.ndarray, in2: np.ndarray) → tuple

Uses normalized cross-correlation to find the coordinates in in2 where in1 can be found.

correlation_coefficient(image_1, image_2)

Create the normalized correlation coefficient (scalar) of two images.

crop_image_around_position_to_size(image, y, x, size)

Crop an image to given size centered at coordinates (y, x).

create_rgb_image(red, green, blue=None)

Merge three single channels into an RGB image.

find_features(image, detector='surf', num_features=1000, hessian_threshold=10, use_latch_descriptor=False)

Find features in of a template inside a larger image.

find_position_of_template_in_image_using_descriptors(template_kps, template_des, image_kps, image_des, template_size)

Find the template in the image using the extracted feature descriptors.

register_images_opencv_features(source, target, detector='surf', use_latch_descriptor=False, perspective=True, affine=False, rigid=False, num_features=1000, hessian_threshold=10, control_image=False)

Register 2 images using image features.

apply_transformation_to_image(image, transformation_type, transformation_matrix, target_height=None, target_width=None)

Apply a transformation to an image.

display_matches(img1, img2, sel_matches, k1, k2, max_matches=None)

Displays the matches on a control image and returns it.

add_border(images: list, border: int, fill_value: int = -1) → list

Add a border to each of the images in a list and sets the border values to a given fill value.

BGR2Gray(image, to_lightness=False)

Convert a BGR image to gray or lightness.
pypocquant.lib.processing.phase_only_correlation(in1: np.ndarray, in2: np.ndarray) → np.ndarray

Calculate phase-only correlation of two numpy arrays.

Parameters:

  • in1 – 2D numpy array.

  • in2 – 2D numpy array.

Returns:

2D np.float64 numpy array.

pypocquant.lib.processing.find_position_in_image_using_phase_corr(in1: np.ndarray, in2: np.ndarray) → tuple

Uses phase-only correlation to find the coordinates in in2 where in1 can be found.

Parameters:

  • in1 – 2D numpy array (must be strictly smaller, i.e. completely contained) in in2.

  • in2 – 2D numpy array.

Returns:

tuple with (y = row, x = column) location of the center of in1 in in2.

Return type:

tuple

pypocquant.lib.processing.find_position_in_image_using_norm_xcorr(in1: np.ndarray, in2: np.ndarray) → tuple

Uses normalized cross-correlation to find the coordinates in in2 where in1 can be found.

Parameters:

  • in1 – 2D numpy array (must be strictly smaller, i.e. completely contained) in in2.

  • in2 – 2D numpy array.

Returns:

tuple with (y = row, x = column) location of the center of in1 in in2.

Return type:

tuple

pypocquant.lib.processing.correlation_coefficient(image_1, image_2)

Create the normalized correlation coefficient (scalar) of two images. :param image_1: np image1 :param image_2: np image2

Returns:

product:

pypocquant.lib.processing.crop_image_around_position_to_size(image, y, x, size)

Crop an image to given size centered at coordinates (y, x).

If the original image is too small, a cropped version will be returned.

Parameters:

  • image – Image to be cropped

  • y – y center coordinate

  • x – x center coordinate

  • size – size of the crop

Returns:

out: Cropped image

pypocquant.lib.processing.create_rgb_image(red, green, blue=None)

Merge three single channels into an RGB image.

Parameters:

red – Red channel

:param green

Green channel

:param blue

Blue channel

Returns:

view: RGB image

pypocquant.lib.processing.find_features(image, detector='surf', num_features=1000, hessian_threshold=10, use_latch_descriptor=False)

Find features in of a template inside a larger image.

Parameters:

  • image –

  • detector –

  • num_features –

  • hessian_threshold –

  • use_latch_descriptor –

Returns:

kp

Returns:

des

pypocquant.lib.processing.find_position_of_template_in_image_using_descriptors(template_kps, template_des, image_kps, image_des, template_size)

Find the template in the image using the extracted feature descriptors.

Parameters:

  • template_kps –

  • template_des –

  • image_kps –

  • image_des –

  • template_size –

Returns:

coordinates

Return type:

tuple

pypocquant.lib.processing.register_images_opencv_features(source, target, detector='surf', use_latch_descriptor=False, perspective=True, affine=False, rigid=False, num_features=1000, hessian_threshold=10, control_image=False)

Register 2 images using image features.

Keyword arguments: :param source: source image to be registered (must be grayscale) :param target: target image (must be grayscale) :param detector: one of “orb”, “kaze”, “akaze”, “brisk”, “surf”, “sift” (default if surf) :param use_latch_descriptor: True to use the new LATCH descriptor (requires openCV 3.1), False to use the

default descriptors provided by the detectors (default is False)

Parameters:

  • num_features – number of features (used only by the “orb” and “sift” detectors, default is 1000)

  • hessian_threshold – threshold of the hessian of the images (used only by the “surf” detector, default is 10)

  • perspective – register the image using a perspective transformation (optional, default=True).

  • affine – register the image using an affine transformation (optional, default=False).

  • rigid – register the image using a rigid transformation (optional, default=False).

  • control_image – set to True to create a quality control image (default is False).

Returns:

results (aligned: aligned image, M : transformation matrix, mask: mask returned by cv2.findHomography()),

Return type:

dict

Returns:

view: quality control image,

Returns:

source_descr: list of source descriptors,

Returns:

target_descr: list of target descriptors).

pypocquant.lib.processing.apply_transformation_to_image(image, transformation_type, transformation_matrix, target_height=None, target_width=None)

Apply a transformation to an image.

Parameters:

  • image – image to be transformed.

  • transformation_type –

    type of transformation. One of:

    ”perspective”: register the image using a perspective transformation. “affine”: register the image using an affine transformation. “rigid”: register the image using a rigid transformation.

  • transformation_matrix –

    transformation matrix, must be:

    ”perspective”: (3x3) “affine”: (2x3) “rigid”: (2x3)

  • target_height – (optional) number of rows of the transformed image. If not set, the transformed image will have the same size as the source image.

  • target_width – (optional) number of columns of the transformed image. If not set, the transformed image will have the same size as the source image.

Returns:

transformed: transformed image.

pypocquant.lib.processing.display_matches(img1, img2, sel_matches, k1, k2, max_matches=None)

Displays the matches on a control image and returns it.

Parameters:

  • img1 – First image

  • img2 – Second image

  • sel_matches – Selected matches

  • k1 –

  • k2 –

  • max_matches –

Returns:

view

pypocquant.lib.processing.add_border(images: list, border: int, fill_value: int = - 1) → list

Add a border to each of the images in a list and sets the border values to a given fill value.

If the fill_value is omitted, the median of all pixel intensities will taken.

Parameters:

  • images (list) – List of images

  • border (int) – Border to be added to image

  • fill_value – (optional) If omitted the median of all pixel intensities will taken.

Returns:

out: List of images with added border.

Return type:

list

pypocquant.lib.processing.BGR2Gray(image, to_lightness=False)

Convert a BGR image to gray or lightness.

Parameters:

  • image – Image to be converted

  • to_lightness – To lightness bool

Returns:

l

Return type:

cv2.Image