stouputils.data_science.data_processing.prosthesis_detection module#

get_brightness_score(
image: ndarray[tuple[int, ...], dtype[Any]],
rect: Any,
percentile: int = 95,
) → float[source]#

Compute brightness score using high-percentile pixel intensity.

get_contrast_score(
image: ndarray[tuple[int, ...], dtype[Any]],
rect: Any,
) → float[source]#

Compute contrast score between object and surrounding background.

get_corners_distance(
rect: Any,
image_shape: tuple[int, int],
) → float[source]#

Compute average distance between rectangle corners and image center.

get_box_overlap_ratio(
box1: Any,
box2: Any,
) → float[source]#

Compute overlap ratio between two bounding boxes with intersection area.

get_fracture_score(
image: ndarray[tuple[int, ...], dtype[Any]],
rect: Any,
padding: int = 20,
) → float[source]#

Compute fracture score based on bone fractures around prosthesis.

keep_bright_enough_parts(
image: ndarray[tuple[int, ...], dtype[Any]],
window_size: int = 101,
invert: bool = False,
debug_level: int = 0,
) → ndarray[tuple[int, ...], dtype[Any]][source]#

Keep only the brightest parts of the image.

For each pixel, if the window around it is brighter than 60% of the brightest pixels in the image, keep it.

Parameters:

  • image (NDArray[Any]) – Image to process.

  • window_size (int) – Size of the window to consider around each pixel.

  • invert (bool) – Instead of keeping the brightest parts, keep the darkest parts.

  • debug_level (int) – Debug level.

Returns:

Processed image with only bright parts preserved.

Return type:

NDArray[Any]