Visual similarity
Image Similarity Checker
Upload two images and see how visually similar they are, using perceptual hashing — not an exact pixel match.
Your files never leave your browser. Nothing is uploaded, stored, or seen by us.
Image A
Image B
How it works
Each image is drawn onto a tiny in-memory canvas, converted to grayscale, and reduced to a 9×9 grid of brightness values. The tool then compares each pixel to its neighbor to build a compact fingerprint (a dHash) — this happens entirely inside your browser. The two fingerprints are compared bit by bit; the fewer bits differ, the higher the similarity score.
Frequently asked questions
- How does the image similarity checker work?
- It computes a perceptual hash (dHash) for each image by shrinking it to a small grayscale grid and comparing the relative brightness of neighboring pixels. Two images with similar structure produce similar hashes, even if one has been resized, recompressed, or lightly cropped — unlike a byte-for-byte file comparison.
- Will this detect images that have been edited or cropped?
- Perceptual hashing is resilient to resizing, recompression, and minor color adjustments, so lightly edited duplicates typically still score high. Heavy cropping, rotation, or watermarking will lower the similarity score because the overall visual structure changes more significantly.
- What similarity percentage counts as a match?
- As a rough guide: 95%+ usually means near-identical images (same photo, different compression). 85–94% suggests the same image with minor edits. 70–84% indicates related but noticeably different images. Below 70% generally means the images are structurally different.
- Are my images uploaded anywhere to compare them?
- No. Both images are hashed directly in your browser using the Canvas API — the actual image files never leave your device.