raygeo.image.preprocess

Functions

compute_adaptive_threshold()

compute_adaptive_threshold(areas: list[int]) -> int

Compute an adaptive area threshold to separate noise from content.

Analyses the distribution of connected component areas and finds the largest gap to determine a threshold that separates noise (small components) from meaningful content.

Parameter	Type	Description
areas	list[int]	Sorted list of component pixel areas.
Returns	int	Adaptive threshold value (minimum area to keep).
Complexity		O(n) where n = number of unique area values

Adaptive threshold from component area distribution

denoise_binary()

denoise_binary(
    binary: numpy.NDArray[numpy.uint8],
) -> numpy.NDArray[numpy.uint8]

Remove small noise components from a binary image using adaptive thresholding.

Computes connected components, finds the largest gap in component area distribution to separate noise from content, and removes small components.

Parameter	Type	Description
binary	numpy.NDArray[numpy.uint8]	2D binary uint8 array (values 0 or 1).
Returns	numpy.NDArray[numpy.uint8]	2D binary uint8 array with noise removed.
Complexity		O(w*h)

Binary image denoised via adaptive thresholding

filter_components()

filter_components(
    binary: numpy.NDArray[numpy.uint8],
    min_area: int,
) -> numpy.NDArray[numpy.uint8]

Remove connected components smaller than min_area.

Uses 8-connectivity for component detection.

Parameter	Type	Description
binary	numpy.NDArray[numpy.uint8]	2D binary uint8 array (values 0 or 1).
min_area	int	Minimum pixel count to keep a component.
Returns	numpy.NDArray[numpy.uint8]	2D binary uint8 array (values 0 or 1).
Complexity		O(w*h)

Component filtering by minimum area

get_component_areas()

get_component_areas(binary: numpy.NDArray[numpy.uint8]) -> list[int]

Compute the pixel area of each connected component.

Uses 8-connectivity. Areas are returned sorted ascending. Background (0-valued pixels) is excluded.

Parameter	Type	Description
binary	numpy.NDArray[numpy.uint8]	2D binary uint8 array (values 0 or 1).
Returns	list[int]	Sorted list of component pixel areas.
Complexity		O(w*h)

Connected component areas sorted ascending

grayscale_to_binary()

grayscale_to_binary(
    gray: numpy.NDArray[numpy.uint8],
    threshold: float = 0.5,
    invert: bool = False,
    auto_threshold: bool = True,
) -> numpy.NDArray[numpy.uint8]

Convert grayscale image to binary using Otsu or fixed threshold.

Pixels at or below the threshold become foreground (1). Uses Otsu's method when auto_threshold is True.

Parameter	Type	Description
gray	numpy.NDArray[numpy.uint8]	2D grayscale uint8 image.
threshold	float = 0.5	Fixed threshold (0.0-1.0), used only if auto_threshold is False.
invert	bool = False	If True, pixels above threshold become foreground.
auto_threshold	bool = True	If True, compute threshold via Otsu's method.
Returns	numpy.NDArray[numpy.uint8]	2D binary uint8 array (values 0 or 1).
Complexity		O(w*h)

Grayscale to binary via Otsu and fixed threshold