Find the Largest Rectangle Inside a Mask

Find the largest axis-aligned rectangle that fits entirely within the foreground region. Useful for cropping to usable content.

Basic usage

bbox = mask.largest_interior_rectangle()
x, y, w, h = bbox

Returns [x, y, width, height] of the largest rectangle.

With aspect ratio constraint

Find the largest rectangle with a specific aspect ratio:

# 16:9 aspect ratio
bbox = mask.largest_interior_rectangle(aspect_ratio=16/9)

The returned rectangle will have exactly that width/height ratio.

Centered on a point

Find the largest rectangle that must contain a specific center point:

center = (320, 240)  # x, y
bbox = mask.largest_interior_rectangle_around(center)

With both constraints

bbox = mask.largest_interior_rectangle_around(
    center=(320, 240),
    aspect_ratio=4/3
)

Use case: optimal crop after warping

After warping an image, the valid region may be non-rectangular. Find the best crop that avoids black borders:

# valid_mask marks pixels that have real data (not border fill)
crop_box = valid_mask.largest_interior_rectangle(aspect_ratio=16/9)

x, y, w, h = crop_box
cropped_image = image[y:y+h, x:x+w]

Use case: undistortion

When removing lens distortion, find the largest usable rectangle:

# valid_region is where the undistortion mapping is defined
inner_crop = valid_region.largest_interior_rectangle()

# Or keep principal point centered
principal_point = (intrinsic_matrix[0, 2], intrinsic_matrix[1, 2])
centered_crop = valid_region.largest_interior_rectangle_around(principal_point)