Checking duplicate images before posting - best way?

[Jake Gittes]

Great topic!

Totally agree with the Image Grabber idea!

I'll just add that from my experience it helps to be as descriptive as possible when posting pics and it also helps to post sets if you can!

Good thread idea london_manuk!

[deepsepia]

Originally Posted by ConstantOgler

This one of many = help threads here that prompt me to ask: This is all about file name matching, no?

No, its about matching the file hash. That's a number mathematically generated from a file's contents-- change the name and the hash stays the same. Each set of bits in a file will produce an effectively unique hash (at a theoretical level, there are cases where two different files might have the same hash, but that's so rare as to be irrelevant).

So you can change the name, and things like Tineye find the image fine . . . because the filename is irrelevant. This is why you'll sometimes see folks add some irrelevant extra file to a zip that they're uploading to a host . . . if a filehost has had a particular file reported for piracy, they'll have its hash on a blacklist. . . adding a even a 20K jpg to the file and zipping them up together produces an upload with a different hash.

[Midnight Air]

Originally Posted by ConstantOgler

This one of many = help threads here that prompt me to ask: This is all about file name matching, no? And worst case, it requires exact name match. (I don't know how fuzzy/flexible these programs are.) How practical is it to think filename will be the same? Filenames are changed constantly by hosts & others. And if it's your own new scan you want to check, name match is meaningless.



Is it doing bitmap file analysis/comparison ("Computer Vision") or is it all about text, i.e., filenames?

It finds similar looking files irrespective of their hash or filenames.
following options: exact files, similar files, ranging from 1 to 100 percent similarity. it scans the bitmap regions, and an option is to keep the results in a database [sqllite].