How is this different from an MD5 checksum?
A checksum changes completely if a single byte changes, so a resaved image looks entirely unrelated. A perceptual hash stays close when the picture still looks the same.
Generate perceptual hashes for an image: average hash, difference hash and colour statistics. Unlike a checksum, these stay similar when an image is resized or recompressed, which is what makes them useful for finding duplicates. It is free, needs no account, and adds no watermark to your result.
Average and difference hashes are computed from a downscaled greyscale version of the image, alongside colour statistics, and returned as JSON. Similar images produce similar hashes.
Where it stops: These are similarity measures, not identity proofs. Heavy cropping or a large colour shift will change the hash substantially.
A checksum changes completely if a single byte changes, so a resaved image looks entirely unrelated. A perceptual hash stays close when the picture still looks the same.
Count the differing bits, known as the Hamming distance. A small distance means the images are visually similar.
Often not. Cropping changes composition significantly, which is exactly what these hashes measure.
Calculate SHA-256, SHA-512 and MD5 checksums for uploaded files without changing their contents.
Inspect image dimensions, format, EXIF fields, camera data, orientation, DPI and embedded metadata as JSON.
Find the single dominant color of an image and export its HEX, RGB and pixel-share information.
Inspect file name, extension, MIME guess, size, timestamps where available and format-specific details for common media.