Find near-duplicate images. Computes perceptual hash distance between two image URLs to detect crops, resizes, re-compressions, and slightly edited copies of the same photo. Use for content moderation, copyright checks, and cleaning image datasets.