dataeval.core.dhash_d4¶
- dataeval.core.dhash_d4(image)¶
Compute orientation-invariant difference hash using gradients.
Computes dhash for all 8 dihedral group transformations (4 rotations × 2 flip states) and returns the lexicographically smallest hash as the canonical representative.
- Parameters:¶
- image : Array3D¶
An image in CxHxW format. Can be a 3D list, or array-like object.
- Returns:¶
Canonical hex string hash invariant to rotation and mirroring, or empty string if image is too small or not spatial data.
- Return type:¶
str
Notes
This hash is invariant to: - 90°, 180°, 270° rotations - Horizontal and vertical flips - Any combination of rotation and flip
The canonical hash is the lexicographically smallest hash among all 8 orientations, ensuring that any orientation of the same image produces the identical hash.
Computation cost is ~8x that of regular dhash.