Image Cropping

Updated 02 Oct 2026

•
What it is
- •
  extracts a spatial region without resampling its retained pixels
•
Coordinates
- •
  for a box $(x,y,w,h)$ , the intended array slice is image[y:y+h, x:x+w].
- •
  array order is row then column, while the box convention gives horizontal position first.
•
Validation
- •
  intersect the box with image bounds and verify positive width and height before saving.
- •
  negative NumPy indices count from the end; they do not mean “clip to zero.”
•
Project
- •
  the face-dataset route crops the largest detected box
- •
  it directly slices the detector's integer coordinates withotu explicitly clipping them to image bounds
•
Distinguish from
- •
  Image Resizing changes the sampling grid
•
Related
- •
  Face Detection

Linked references 3