Axis Aligned

Axis-Aligned LIR

The Axis-Aligned LIR algorithm solves the Largest Inscribed Rectangle problem under a fixed-axis (axis-aligned) orientation constraint. Unlike other LIRiAP families, this provides an exact solution at vertex-coordinate precision.

Overview

• Exactness: Guaranteed optimal rectangle for the given axis orientation
• Complexity: O(n²) in polygon vertex count
• Orientation: Fixed axis (or rotated via AXIS_ANGLE parameter)
• Supports: Convex, concave, with holes, without holes

Algorithm Flow

flowchart TD
    A[Input polygon] --> B[Detect polygon type]
    B --> C{Convex, no holes?}
    C -- Yes --> D[Alt/Amenta solver]
    C -- No --> E[Daniels et al solver]
    D --> F[Epsilon-inset certification]
    E --> F
    F --> G[Apply max_ratio constraint]
    G --> H[Optional axis rotation]
    H --> I[Output exact rectangle]

Loading

Exact Solvers

The algorithm dispatches to one of four exact solvers based on polygon topology:

Polygon Type	Solver	Complexity	Guarantee
Convex, no holes	Alt/Amenta	O(n²)	Exact, vertex-pair enumeration
Convex with holes	Daniels et al.	O(n²)	Exact, vertex-coordinate grid
Concave, no holes	Daniels et al.	O(n²)	Exact, vertex-coordinate grid
Concave with holes	Daniels et al.	O(n²)	Exact, vertex-coordinate grid

Alt/Amenta Solver (Convex No Holes)

For convex polygons, the optimal axis-aligned rectangle has its bottom and top sides tangent to the polygon boundary, with each contact at a vertex y-coordinate.

• Enumerate all O(n²) pairs of vertex y-coordinates as (y_lo, y_hi) candidates
• Compute horizontal polygon chord intersection at each y level
• Take maximum-width rectangle

Reference: Alt, H. et al. (1987); Amenta, N. (1994)

Daniels et al. Solver (All Other Cases)

Daniels et al. (1997) proved that the largest axis-aligned rectangle inscribed in a simple polygon always has at least two sides determined by vertex x- or y-coordinates.

• Build vertex-coordinate grid from all unique vertex x/y values
• Create cell-centre PIP (point-in-polygon) mask
• Run Largest Rectangle in Histogram (LRH) stack sweep row-by-row
• Returns exact answer at vertex-coordinate precision

References:

• Daniels, K., Milenkovic, V., Roth, D. (1997). Finding the largest axis-aligned rectangle in a polygon. Proc. 13th CCCG.
• Klingel, E. (1986). Finding the largest rectangle in a polygon (LRH).

Vertex-Grid Midpoint Augmentation

The solver augments raw vertex coordinates with midpoints between consecutive unique values on each axis. This doubles grid density and handles cases where optimal rectangle interior lies between two vertex coordinates (e.g., right triangles, regular hexagons).

Without augmentation, the single cell center can land exactly on the polygon boundary, causing Shapely's strict contains to return False for a cell that is geometrically inside.

Rotation Axis

The AXIS_ANGLE parameter (degrees) rotates the frame in which "axis-aligned" is interpreted:

• Rotate polygon by -axis_angle before solving
• Solve exact axis-aligned problem in rotated frame
• Rotate result rectangle back by +axis_angle

This provides exact solutions for tilted-but-rectangular inscribed rectangles.

Parameters

Parameter	Description	Default
AXIS_ANGLE	Rotation of axis-aligned frame (0 = horizontal)	0.0
GRID_FINE	Fallback grid resolution (used when vertex density > 500)	120
MAX_RATIO	Maximum aspect ratio (long:short), 0 = unlimited	1.6
ALWAYS_RETURN	Return best-effort if epsilon-inset fails	True
USE_BUFFER	Apply containment buffer after certification	False
BUFFER_VALUE	Buffer distance in map units (negative = inward)	-0.5

Output Fields

Field	Type	Description
feat_id	int	Source feature ID
area	double	Rectangle area in CRS map units
axis_angle	double	Axis angle used (echoes input parameter)
poly_type	string	convex_no_holes / convex_with_holes / concave_no_holes / concave_with_holes
ratio	double	Actual long:short aspect ratio
best_effort	int	1 if result from shrink fallback, 0 otherwise

Semantics

• Exact: Rectangle sides snap exactly to vertex coordinates (or midpoint-augmented grid lines)
• Containment: Guaranteed up to floating-point epsilon via light-weight inset certification
• No expansion needed: Unlike BCRS, exact vertex-coordinate precision means no SDF expansion is required

Performance

The algorithm is O(n²) in vertex count, making it suitable for polygons with typical vertex counts (<1000). For very high-density polygons, the fallback grid solver is used (>500 unique vertices triggers grid fallback).

References

1. Alt, H., Hagerup, T., Mehlhorn, K., Preparata, F.P. (1987). Deterministic simulation of idealized parallel computers. Information and Computation.
2. Amenta, N. (1994). A short proof of an interesting heuristic result. Proc. 5th ACM-SIAM SODA.
3. Daniels, K., Milenkovic, V., Roth, D. (1997). Finding the largest axis-aligned rectangle in a polygon. Proc. 13th Canadian Conf. Computational Geometry.
4. Klingel, E. (1986). Finding the largest rectangle in a polygon.