ndimage

ndimage is a subpackage of SciPy that provides a suite of image processing and analysis tools for multi-dimensional arrays. It is designed to work with NumPy arrays and supports images of arbitrary dimensionality, with common use in 2D and 3D imaging tasks.

The module includes routines for filtering and smoothing (gaussian_filter, median_filter, uniform_filter, minimum_filter, maximum_filter), convolution and correlation (convolve, correlate), interpolation and coordinate mapping (map_coordinates, zoom, shift), and geometric transformations (affine_transform, rotate). It also offers a set of morphological operations for binary and grayscale images (binary_erosion, binary_dilation, binary_opening, binary_closing, distance_transform_edt) as well as tools for segmentation and measurement such as labeling of connected components (label) and region-wise statistics (center_of_mass, sum, mean).

ndimage operates on NumPy arrays in N dimensions and supports various boundary handling modes (such as reflect, nearest, and constant) to control edge behavior during filtering and transformation. It is implemented with performance considerations in mind, with core routines often implemented in C and exposed through a Python interface for usability and integration with NumPy workflows.

As part of the SciPy ecosystem, ndimage is widely used in image processing, computer vision, medical imaging, microscopy, and other scientific computing domains that require practicalND image analysis capabilities. It serves as a foundational toolkit for preprocessing, feature extraction, segmentation, and quantitative analysis of multidimensional image data, complementing higher-level libraries in the SciPy/NumPy stack.