nextcv.image.stitching ⚓︎

Generalized image stitching framework.

Simple, extensible architecture for stitching images from multiple cameras.

CLASS	DESCRIPTION
`AdditiveCompensator`	Sequential additive bias correction.
`ExposureCompensator`	Abstract base class for exposure compensation strategies.
`HorizontalStitcher`	Stitch cameras arranged horizontally (left-to-right).
`ImageStitcher`	Abstract base class for image stitching.
`LeftRightStitcher`	Stitch cameras arranged horizontally (left-to-right).
`NoOpCompensator`	No compensation - returns images unchanged.
`Rect`	Image region defined by top-left corner and dimensions.
`Tile`	A tile within a canvas for image stitching.

nextcv.image.stitching.AdditiveCompensator ⚓︎

Bases: ExposureCompensator

Sequential additive bias correction.

Computes median intensity difference in overlap regions and applies additive corrections. First image is used as reference. Each subsequent image is corrected to match the already-corrected previous image.

Good for small-to-medium brightness differences in thermal cameras.

METHOD	DESCRIPTION
`__call__`	Compute and apply additive bias corrections.

call ⚓︎

__call__(tiles: List[Tile], images: List[ndarray]) -> List[ndarray]

Compute and apply additive bias corrections.

_compute_bias_between_tiles `staticmethod` ⚓︎

_compute_bias_between_tiles(tile_i: Tile, tile_j: Tile, img_i: ndarray, img_j: ndarray) -> float

Compute intensity bias between two overlapping tiles.

PARAMETER	DESCRIPTION
`tile_i`	First tile (reference) TYPE: `Tile`
`tile_j`	Second tile (to be corrected) TYPE: `Tile`
`img_i`	Warped image from tile_i TYPE: `ndarray`
`img_j`	Warped image from tile_j TYPE: `ndarray`

RETURNS	DESCRIPTION
`float`	Bias to add to img_j to match img_i (float scalar)

nextcv.image.stitching.ExposureCompensator ⚓︎

Bases: ABC

Abstract base class for exposure compensation strategies.

Compensators compute and apply corrections to warped images to minimize intensity differences in overlap regions.

METHOD	DESCRIPTION
`__call__`	Compute and apply corrections to warped images.

call `abstractmethod` ⚓︎

__call__(tiles: List[Tile], warped_images: List[ndarray]) -> List[ndarray]

Compute and apply corrections to warped images.

PARAMETER	DESCRIPTION
`tiles`	List of tiles with rect, mask, and weights TYPE: `List[Tile]`
`warped_images`	List of warped images (one per tile) TYPE: `List[ndarray]`

RETURNS	DESCRIPTION
`List[ndarray]`	List of corrected images

nextcv.image.stitching.HorizontalStitcher ⚓︎

HorizontalStitcher(cameras: List[CameraType], compensator: Optional[ExposureCompensator] = None)

Bases: ImageStitcher[PinholeCamera]

Stitch cameras arranged horizontally (left-to-right).

nextcv.image.stitching.ImageStitcher ⚓︎

ImageStitcher(cameras: List[CameraType], compensator: Optional[ExposureCompensator] = None)

Bases: ABC, Generic[CameraType]

Abstract base class for image stitching.

PARAMETER	DESCRIPTION
`cameras`	List of cameras to stitch TYPE: `List[CameraType]`
`compensator`	Exposure compensator to use. Defaults to NoOpCompensator() TYPE: `Optional[ExposureCompensator]` DEFAULT: `None`

METHOD	DESCRIPTION
`__call__`	Stitch images into panorama.
`create_warp_tiles`	Create normalized warp tiles for all cameras.
`stitch`	Stitch images into the virtual camera.

call ⚓︎

__call__(images: List[ndarray]) -> ndarray

Stitch images into panorama.

_create_virtual_cam `abstractmethod` ⚓︎

_create_virtual_cam() -> CameraType

Create virtual camera defining output coordinate system.

_sanity_checks ⚓︎

_sanity_checks(images: List[ndarray]) -> None

Sanity checks for the images.

create_warp_tiles `staticmethod` ⚓︎

create_warp_tiles(cameras: List[CameraType], canvas: CameraType) -> List[Tile]

Create normalized warp tiles for all cameras.

stitch ⚓︎

stitch(images: List[ndarray]) -> ndarray

Stitch images into the virtual camera.

nextcv.image.stitching.LeftRightStitcher ⚓︎

LeftRightStitcher(left_camera: PinholeCamera, right_camera: PinholeCamera)

Bases: HorizontalStitcher

Stitch cameras arranged horizontally (left-to-right).

nextcv.image.stitching.NoOpCompensator ⚓︎

Bases: ExposureCompensator

No compensation - returns images unchanged.

Fastest option. Use when images are already well-balanced or when exposure differences are negligible.

METHOD	DESCRIPTION
`__call__`	Return images unchanged.

call ⚓︎

__call__(tiles: List[Tile], warped_images: List[ndarray]) -> List[ndarray]

Return images unchanged.

nextcv.image.stitching.Rect `dataclass` ⚓︎

Rect(x: int, y: int, w: int, h: int)

Image region defined by top-left corner and dimensions.

PARAMETER	DESCRIPTION
`x`	TYPE: `int`
`y`	TYPE: `int`
`w`	TYPE: `int`
`h`	TYPE: `int`

METHOD	DESCRIPTION
`clamp_to`	Clamp region to bounds, return None if no overlap.
`intersect`	Compute intersection with another rectangle.
`numpy_slices`	Return (y_slice, x_slice) for numpy indexing with optional offset.

ATTRIBUTE	DESCRIPTION
`s`	Shorthand for numpy_slices() - returns (y_slice, x_slice). TYPE: `Tuple[slice, slice]`

s `property` ⚓︎

s: Tuple[slice, slice]

Shorthand for numpy_slices() - returns (y_slice, x_slice).

Usage

arr[rect.s] = value

clamp_to ⚓︎

clamp_to(max_w: int, max_h: int) -> Optional[Rect]

Clamp region to bounds, return None if no overlap.

intersect ⚓︎

intersect(other: Rect) -> Optional[Rect]

Compute intersection with another rectangle.

numpy_slices ⚓︎

numpy_slices(offset_x: int = 0, offset_y: int = 0) -> Tuple[slice, slice]

Return (y_slice, x_slice) for numpy indexing with optional offset.

nextcv.image.stitching.Tile `dataclass` ⚓︎

Tile(rect: Rect, maps: Tuple[ndarray, ndarray], mask: ndarray, weights: ndarray)

A tile within a canvas for image stitching.

PARAMETER	DESCRIPTION
`rect`	TYPE: `Rect`
`maps`	TYPE: `Tuple[ndarray, ndarray]`
`mask`	TYPE: `ndarray`
`weights`	TYPE: `ndarray`

METHOD	DESCRIPTION
`from_camera_pair`	Create a warp tile from a camera.
`update_weights`	Update the weights.
`warp_image`	Warp source image to this tile.

from_camera_pair `classmethod` ⚓︎

from_camera_pair(camera: CameraType, canvas: CameraType) -> Optional[Tile]

Create a warp tile from a camera.

update_weights ⚓︎

update_weights(weights: ndarray) -> Tile

Update the weights.

warp_image ⚓︎

warp_image(src_img: ndarray) -> ndarray

Warp source image to this tile.

nextcv.image.stitching ⚓︎

nextcv.image.stitching.AdditiveCompensator ⚓︎

__call__ ⚓︎

_compute_bias_between_tiles staticmethod ⚓︎

nextcv.image.stitching.ExposureCompensator ⚓︎

__call__ abstractmethod ⚓︎

nextcv.image.stitching.HorizontalStitcher ⚓︎

nextcv.image.stitching.ImageStitcher ⚓︎

__call__ ⚓︎

_create_virtual_cam abstractmethod ⚓︎

_sanity_checks ⚓︎

create_warp_tiles staticmethod ⚓︎

stitch ⚓︎

nextcv.image.stitching.LeftRightStitcher ⚓︎

nextcv.image.stitching.NoOpCompensator ⚓︎

__call__ ⚓︎

nextcv.image.stitching.Rect dataclass ⚓︎

s property ⚓︎

clamp_to ⚓︎

intersect ⚓︎

numpy_slices ⚓︎

nextcv.image.stitching.Tile dataclass ⚓︎

from_camera_pair classmethod ⚓︎

update_weights ⚓︎

warp_image ⚓︎

call ⚓︎

_compute_bias_between_tiles `staticmethod` ⚓︎

call `abstractmethod` ⚓︎

call ⚓︎

_create_virtual_cam `abstractmethod` ⚓︎

create_warp_tiles `staticmethod` ⚓︎

call ⚓︎

nextcv.image.stitching.Rect `dataclass` ⚓︎

s `property` ⚓︎

nextcv.image.stitching.Tile `dataclass` ⚓︎

from_camera_pair `classmethod` ⚓︎