API

Stitches a 2D grid of images into a mosaic

stitch2d.mosaic

Reads and stitches images from a 2D grid into a mosaic

class stitch2d.mosaic.Mosaic(path_or_tiles, tile_class=None)

Bases: object

Stitches a mosaic from a list of tiles

grid

tiles arranged into a grid

Type:

list

shape

shape of the mosaic as (height, width[, channels])

Type:

tuple

size

number of tiles in the mosaic

Type:

tuple

tile_class

class to use for tiles in the mosaic

Type:

class

num_cores = 1

Number of cores to use when processing images

Type:

int

__init__(path_or_tiles, tile_class=None)

Initializes a mosaic from a list of tiles

Parameters:

• path_or_tiles (str or list-like) – either the path to a directory containing tiles, a list of Tiles, or a list of strings or arrays that can be used to create a Tile

• tile_class (class) – class to use for tiles in the mosaic. Defaults to OpenCVTile.

property tiles

Gets a flattened list of all tiles in grid order

property placed

Calculates number of tiles that have been placed in the mosaic

property params

Summarizes parameters needed to stitch mosaic

property detector

Gets a copy of the detector used to align tiles in the mosaic

property matcher

Gets a copy of the matcher used to align tiles in the mosaic

Only defined if using OpenCV.

property pool

Returns a shared joblib pool, creating it if needed

placeholder(tile=None, fill_value=0)

Creates a placeholder tile to fill in gaps in the mosaic

Parameters:

• tile (Tile) – tile to base the placeholder on. If not given, uses the first tile in the tiles property.

• fill_value (float or int) – fill value

Returns:

placeholder tile filled with provided value

Return type:

Tile

bounds()

Calculates bounds of the mosaic comprising the placed tiles

Returns:

bounds of tile as (y1, x1, y2, x2)

Return type:

tuple

copy()

Creates a copy of the mosaic based on the grid

Using the grid instead of the list of tiles allows the grid-building step to be skipped when the copy is initialized.

Returns:

copy of the mosaic

Return type:

Mosaic

save_params(path='params.json')

Saves coordinates for placed tiles

Parameters:

path (str) – path to the JSON file

load_params(path_or_obj='params.json')

Loads coordinates for placed tiles at full scale

Parameters:

path_or_obj (str or dict) – path to the JSON file or param dict from another mosaic

Raises:

• FileNotFoundError – thrown if input is path and path not found

• ValueError – thrown if JSON can’t be decoded or does not match this mosaic

reset_tiles()

Reloads tiles at their full resolution

resize(size_or_shape, *args, **kwargs)

Rescales all tiles in the mosaic using size or shape

Parameters:

• size_or_shape (int or tuple of ints) – size in megapixels or shape of resized image

• *args – any argument accepted by the resize function used by the tile class

• **kwargs – any keyword argument accepted by the resize function used by the tile class

downsample(size_or_shape, *args, **kwargs)

Downsamples all tiles in the mosaic using the given size or shape

Parameters:

• size_or_shape (int or tuple of ints) – size in megapixels or shape of resized image

• *args – any argument accepted by the resize function used by the tile class

• **kwargs – any keyword argument accepted by the resize function used by the tile class

detect_and_extract()

Detects and extracts features in tiles

align(origin=None, limit=None, **kwargs)

Builds a mosaic by checking each tile against all others

Parameters:

• origin (Tile) – the tile around which to build the mosaic. If not given, method will select the tile with the largest number of features.

• limit (int) – the number of tiles that must be successfully placed before the method finishes. If not given, the method will continue until it runs out of adjacent tiles with matching features. Setting a limit allows a decent mosaic to be created quickly.

• kwargs – any keyword argument accepted by the align_to method on the Tiles comprising this mosaic

build_out(*args, **kwargs)

Warns user that build_out is not implemented in Mosaic class

Use StructuredMosaic instead to get this functionality.

smooth_seams(origin=None)

Smooths intensities at seams between tiles

Parameters:

origin (Tile) – starting tile

stitch(channel_order=None)

Stitches mosaic using either placed tiles or row/col of tiles

Parameters:

channel_order (str) – order of the three color channels in the stitched array, for example, RGB. Uses the backend order if not given, which can give unexpected results (for example, OpenCV uses BGR).

Return type:

numpy.ndarray

save(path)

Saves mosaic to path

Parameters:

path (str) – file path

show(*args, **kwargs)

Shows the mosaic

class stitch2d.mosaic.StructuredMosaic(path_or_tiles, tile_class=None, dim=None, origin='upper left', direction='horizontal', pattern='raster')

Bases: Mosaic

Stitches a mosaic from a list of tiles with a known structure

__init__(path_or_tiles, tile_class=None, dim=None, origin='upper left', direction='horizontal', pattern='raster')

Initializes a structured mosaic from a list of tiles

Parameters:

• path_or_tiles (str or list-like) – either the path to a directory containing tiles, a list of Tiles, or a list of strings or arrays that can be used to create a Tile

• tile_class (class) – class to use for tiles in the mosaic. Defaults to OpenCVTile.

• dim (tuple or int) – either the shape of the mosaic as (height, width) or the number of tiles in the direction traversed first, that is, the number of columns (if horizontal) or number of rows (if vertical)

• origin (str) – the position of the first tile in the mosaic. One of “upper left”, “upper right”, “lower left”, or “lower right”.

• direction (str) – direction to traverse first when building the mosaic. Either “horizontal” or “vertical”.

• pattern (str) – whether the grid is rastered or snaked. Either “raster” or “snake”.

align(origin=None, limit=None, **kwargs)

Builds a mosaic outward from a single tile using feature matching

Parameters:

• origin (Tile) – the tile around which to build the mosaic. If not given, method will select a tile near the center of the mosaic.

• limit (int) – the number of tiles that must be successfully placed before the method finishes. If not given, the method will continue until it runs out of adjacent tiles with matching features. Setting a limit allows a decent mosaic to be created quickly.

• kwargs – any keyword argument accepted by the align_to method on the Tiles comprising this mosaic

build_out(from_placed=True, offsets=None)

Builds out from already placed tiles using the given offset

Used to complete mosaics that include tiles that were not placed when the mosaic was built, either because the user assigned a limit or because the feature matching algorithm failed to find a home for them.

Parameters:

• from_placed (bool) – if True, unplaced tiles will be tacked onto already placed tiles using the given offsets. If False, a new mosaic will be calculated from scratch using the given offsets.

• offsets (tuple) – offsets between adjacent tiles as dy_row, dx_row, dy_col, dx_col. If not given, the method will estimate the offsets if any tiles have been placed or will ignore offsets if not.

stitch2d.mosaic.build_grid(items, dim, origin='upper left', direction='horizontal', pattern='raster', fill_value=None)

Builds a grid from a list

Parameters:

• items (list) – list to convert to a grid

• dim (tuple or int) – either the shape of the mosaic as (height, width) or the number of tiles in the direction traversed first, that is, the number of columns (if horizontal) or number of rows (if vertical)

• origin (str) – the position of the first tile in the mosaic. One of “upper left”, “upper right”, “lower left”, or “lower right”.

• direction (str) – direction to traverse first when building the mosaic. Either “horizontal” or “vertical”.

• pattern (str) – whether the grid is a raster or snake

• fill_value – value used to fill missing items in a ragged grid

Returns:

List of rows in the grid

Return type:

list

stitch2d.mosaic.create_mosaic(path_or_tiles, tile_class=None, dim=None, origin=None, direction=None, pattern=None)

Creates a mosaic

See StructuredMosaic for available parameters.

Returns:

tiles as either a structured or unstructured mosaic

Return type:

Mosaic or StructuredMosaic

stitch2d.mosaic.is_grid(items)

Tests if an iterable looks like a grid

Parameters:

items (list-like) – list of items

Returns:

True if tiles look like a grid, False if not

Return type:

bool

stitch2d.tile

Reads and helps place a single image from a 2D grid

class stitch2d.tile.Tile(data, detector='sift')

Bases: object

An image tile in a mosaic

source

the original data used to created the tile. Either the path to an image file or an array with 1, 3, or 4 channels.

Type:

str or array-like

imdata

image data

Type:

numpy.ndarray

id

a UUID uniquely identifying the tile

Type:

str

row

the index of the row where the tile appears in the mosaic

Type:

int

col

the index of the column where the tile appears in the mosaic

Type:

int

y

the y coordinate of the image within the mosaic

Type:

float

x

the x coordinate of the image within the mosaic

Type:

float

channel_order

the order of the three color channels in the image, e.g., RGB or BGR

Type:

str

scale

the current scale of the tile relative to the original image

Type:

float

features_detected

whether any features were detected in this image

Type:

bool

descriptors

list of descriptors found in this image

Type:

numpy.ndarray

keypoints

list of coordinates of descriptors found in this image

Type:

numpy.ndarray

detectors = {}

maps strings to a subclass-specific feature detector

Type:

dict

matchers = {}

maps strings to a subclass-specific feature matcher

Type:

dict

__init__(data, detector='sift')

Initializes a mosaic from a list of tiles

Parameters:

• data (str or numpy.ndarray) – path to an image file or an array of image data

• detector (str) – name of the detector used to find/extract features. Currently only sift is supported.

property detector

Gets the detector used to align this tile to another tile

property matcher

Gets the matcher used to align this tile to another tile

property height

Gets the height of the image in pixels

property width

Gets the width of the image in pixels

property channels

Gets the number of channels in the image

property channel_axis

Gets the index where channel info is stored

property dtype

Gets the dtype of the image

property shape

Gets the shape of the image

property size

Gets the size of the image

property mp

Gets the size of the image in megapixels

property placed

Whether the tile has been assigned coordinates in the mosaic

load_imdata()

Loads copy of source data

Returns:

copy of source data

Return type:

numpy.ndarray

copy()

Creates a copy of the tile

Parameters:

grid (list of lists) – grid from the mosaic containing the tile

Returns:

copy of the tile

Return type:

Mosaic

bounds(as_int=False)

Calculates the position of the tile within the mosaic

Parameters:

as_int (bool) – whether bounds are converted to integers before returning

Returns:

bounds of the image in the mosaic coordinate system as (y1, x1, y2, x2)

Return type:

tuple

neighbors()

Finds adjacent tiles

Parameters:

• y (int) – row index

• x (int) – column index

Returns:

neighboring tiles keyed to direction (top, right, bottom, left)

Return type:

dict

convert_mosaic_coords(y1, x1, y2, x2)

Converts mosaic coordinates to image coordinates

Returns:

mosaic coordinates translated to image coordinates

Return type:

tuple

update(other)

Updates attributes to match another tile

Parameters:

other (Tile) – a tile with attributes to copy over to this one

crop(box, convert_mosaic_coords=True)

Crops tile to the given box

Parameters:

• box (tuple) – box to crop to as (y1, x1, y2, x2)

• convert_mosaic_coords (bool) – whether to convert the given coordinates from mosaic to image coordinates

Returns:

image data cropped to the given box

Return type:

numpy.ndarray

intersection(other)

Finds the intersection between two placed tiles

Parameters:

other (Tile) – an adjacent tile that has already been placed in the mosaic

Returns:

the overlapping portion of both tiles

Return type:

tuple of Tile

intersects(other)

Tests if two placed tiles intersect

Parameters:

other (Tile) – an adjacent tile that has already been placed in the mosaic

Returns:

True if tiles intersect, False otherwise

Return type:

bool

reset()

Restores original image and resets coordinate and feature attrs

Returns:

the original tile updated to restore the original image data

Return type:

Tile

match_gamma_to(other)

Scales intensity to match intersecting region of another tile

Parameters:

other (Tile) – a tile that intersects this one

Returns:

the original tile with its intensity modified

Return type:

Tile

draw(others=None)

Creates an image from the provided tiles

Parameters:

others (list of Tiles) – a list of tiles to include in the new image. Only tiles that have been placed will be included.

Returns:

an image including all provided tiles

Return type:

numpy.ndarray

save(path, others=None)

Saves an image created from the provided tiles

Parameters:

• path (str) – file path

• others (list of Tiles) – a list of tiles to include in the new image. Only tiles that have been placed will be included.

show(others=None)

Shows an image created from the provided tiles

Parameters:

others (list of Tiles) – a list of tiles to include in the new image. Only tiles that have been placed will be included.

gray()

Returns copy of image converted to grayscale

Returns:

grayscale version of the original iamge

Return type:

numpy.ndarray

resize(size_or_shape, *args, **kwargs)

Resizes image to a given size or shape

Parameters:

• size_or_shape (float, int, or tuple of ints) – size in megapixels or shape of resized image

• *args – any argument accepted by the resize function used by the subclass

• **kwargs – any keyword argument accepted by the resize function used by the subclass

downsample(size_or_shape, *args, **kwargs)

Downsamples image to a given size or shape if smaller than original

Parameters:

• size_or_shape (float, int, or tuple of ints) – size in megapixels or shape of resized image as (height, width)

• *args – any argument accepted by the resize method of the subclass

• **kwargs – any keyword argument accepted by the resize method of the subclass

Returns:

the original tile downsampled to the given size or shape

Return type:

Tile

prep_imdata()

Returns a copy of the tile data suitable for feature detection

Users may wish to create a subclass with a custom version of this method, for example, to scale intensities, enhance contrast, or select image data from an array that include additional bands.

Returns:

copy of image data

Return type:

numpy.ndarray

detect_and_extract(*args, **kwargs)

Detects and extracts features within the tile

Parameters:

• *args – any argument accepted by the feature detection method on the detector

• **kwargs – any keyword argument accepted by the feature detection method on the detector

align_to(other, **kwargs)

Aligns tile to another, already placed tile

Parameters:

other (Tile) – a tile that has already been placed in the mosaic

static backend_save(path, im)

Saves image to path using the tile backend

Parameters:

• path (str) – file path

• im (numpy.ndarray) – image data

static backend_show(im)

Shows an image using the tile backend

Parameters:

im (numpy.ndarray) – image data

class stitch2d.tile.OpenCVTile(data, detector='sift', matcher='flann')

Bases: Tile

An image tile in a mosaic loaded and manipulated using OpenCV

See Tile for available attributes.

detectors = {'sift': <stitch2d.tile._DefaultInstance object>}

maps strings to a subclass-specific feature detector

Type:

dict

matchers = {'bf': <stitch2d.tile._DefaultInstance object>, 'flann': <stitch2d.tile._DefaultInstance object>}

maps strings to a subclass-specific feature matcher

Type:

dict

__init__(data, detector='sift', matcher='flann')

Initializes a mosaic from a list of tiles

Parameters:

• data (str or numpy.ndarray) – path to an image file or an array of image data

• detector (str) – name of the detector used to find/extract features. Currently only sift is supported.

load_imdata()

Loads copy of source data

Returns:

copy of source data

Return type:

numpy.ndarray

gray()

Returns copy of image converted to grayscale

Returns:

grayscale version of the original iamge

Return type:

numpy.ndarray

resize(size_or_shape, *args, **kwargs)

Resizes image to a given size or shape

Parameters:

• size_or_shape (float, int, or tuple of ints) – size in megapixels or shape of resized image as (height, width)

• *args – any argument accepted by cv.resize

• **kwargs – any keyword argument accepted by cv.resize

Returns:

the original tile resized to the given size or shape

Return type:

OpenCVTile

prep_imdata()

Returns a copy of the tile data suitable for feature detection

The built-in version of this method checks if the imdata attribute is an 8-bit array, returning a copy if so. Otherwise, it rescales the intensities and returns an 8-bit copy of the array. The conversion is simplistic, and users may prefer to create a subclass with a custom version of this method instead, for example, to scale intensities, enhance contrast, or select image data from an array that include additional bands.

Returns:

copy of image data as an 8-bit array

Return type:

numpy.ndarray

detect_and_extract(*args, **kwargs)

Detects and extracts features within the tile

Parameters:

• *args – any argument accepted by the detect_and_extract method on the detector

• **kwargs – any keyword argument accepted by the detect_and_extract method on the detector

Returns:

the original tile updated with features and keypoints

Return type:

OpenCVTile

align_to(other, **kwargs)

Aligns tile to another, already placed tile

Parameters:

other (Tile) – a tile that has already been placed in the mosaic

Returns:

the original tile updated with x and y coordinates

Return type:

OpenCVTile

static backend_save(path, im)

Saves image to path using OpenCV

Parameters:

• path (str) – file path

• im (numpy.ndarray) – image data

static backend_show(im, title='OpenCV Image')

Shows an image using OpenCV

Parameters:

im (numpy.ndarray) – image data

class stitch2d.tile.ScikitImageTile(data, detector='sift')

Bases: Tile

An image tile in a mosaic loaded and manipulated using scikit-image

See Tile for available attributes.

detectors = {'sift': <stitch2d.tile._DefaultInstance object>}

maps strings to a subclass-specific feature detector

Type:

dict

__init__(data, detector='sift')

Initializes a mosaic from a list of tiles

Parameters:

• data (str or numpy.ndarray) – path to an image file or an array of image data

• detector (str) – name of the detector used to find/extract features. Currently only sift is supported.

load_imdata()

Loads copy of source data

Returns:

copy of source data

Return type:

numpy.ndarray

gray()

Returns copy of image converted to grayscale

Returns:

grayscale version of the original iamge

Return type:

numpy.ndarray

resize(size_or_shape, *args, **kwargs)

Resizes image to a given size or shape

Parameters:

• size_or_shape (float, int, or tuple of ints) – size in megapixels or shape of resized image

• *args – any argument accepted by skimage.transform.resize

• **kwargs – any keyword argument accepted by skimage.transform.resize

Returns:

the original tile resized to the given size of shape

Return type:

ScikitImageTile

detect_and_extract(*args, **kwargs)

Detects and extracts features within the tile

Parameters:

• *args – any argument accepted by the detect_and_extract method on the detector

• **kwargs – any keyword argument accepted by the detect_and_extract method on the detector

Returns:

the original tile updated with features and keypoints

Return type:

ScikitImageTile

align_to(other, **kwargs)

Aligns tile to another, already placed tile

Parameters:

other (Tile) – a tile that has already been placed in the mosaic

Returns:

the original tile updated with x and y coordinates

Return type:

ScikitImageTile

static backend_save(path, im)

Saves image to path using skimage

Parameters:

• path (str) – file path

• im (numpy.ndarray) – image data

static backend_show(im)

Shows an image using skimage

Parameters:

im (numpy.ndarray) – image data