Boxa

Boxa is a data structure in the Leptonica image processing library used to store a collection of rectangular regions, commonly called bounding boxes. In code and documentation the type is usually written as BOXA, and it is frequently encountered in OCR systems such as Tesseract, where it manages sets of boxes corresponding to detected characters, words, or other regions.

Each element in a BOXA is a BOX, a rectangle defined by its coordinates within an image. A BOXA also maintains metadata such as the number of boxes and the allocated capacity, enabling dynamic growth as boxes are added or removed.

Leptonica provides a suite of functions to manipulate BOXA objects. Typical operations include creating a new BOXA (boxaCreate), copying (boxaCopy), destroying (boxaDestroy), and querying or setting individual boxes (boxaGetBox, boxaSetBox). Boxes can be appended (boxaAddBox) or inserted, removed, and sorted (boxaSort) by coordinates. Utility routines exist to convert between BOXA and other representations or to extract arrays of coordinates.

BOXA is widely used in OCR pipelines to hold bounding boxes for connected components, characters, or words, and to facilitate layout analysis, region-of-interest selection, and training data generation. It provides an efficient, flexible container for multi-box annotations that can be transformed and analyzed in bulk.

See also Leptonica, Tesseract OCR, bounding box, and box data structures.