dataeval.core.phash¶

dataeval.core.phash(image)¶

Compute perceptual hash using Discrete Cosine Transform (DCT).

Resizes image to a square NxN using Lanczos algorithm where N is 32x32 or the largest multiple of 8 smaller than input dimensions. The resampled image is compressed using DCT and the lowest frequency component is encoded as a bit array of greater or less than median value.

Parameters:¶

image : Array3D¶

An image in CxHxW format. Can be a 3D list, or array-like object.

Returns:¶

Hex string hash of the image, or empty string if image is too small or not spatial data.

Return type:¶

str

Notes

DCT-based hashing (pHash) is robust to: - Scaling and resizing - Minor color adjustments - Compression artifacts

It captures frequency information, making it effective for detecting images that have been resized or slightly modified.