How is it different from, e.g. hashing or fingerprinting an image a la tineye?
Does your system require metadata along with the image? If so, what about once an image escapes the metadata?
Does your system require metadata along with the image? If so, what about once an image escapes the metadata?
This allows Mediachain to always point back to information about an image, even if its metadata is stripped out as it goes viral, for example. Near duplicate image detection is quite far along and works well for images that are cropped, distorted, etc.
More on our approach in this post: https://blog.mediachain.io/perceptual-resolution-9c00ad5ca55...
Details on the implementation in this RFC: https://github.com/mediachain/mediachain/blob/master/rfc/med...