Picter’s Image Fingerprint

How Picter is paving the path towards simplified image file management

Not only have we explored how to simplify file management, we’ve also delved deeper into a topic most important to photographers: digital rights. Our research has focused on this: how could we use blockchain technology to overall improve digital rights management for photographers?

In this blog post, we will share some insights into the research and experimentation we’ve done so far with blockchain technology. We know blockchain is still fairly new and unfamiliar to some people, so we recommend you also read our blockchain introduction blog post that gives an overview on what blockchain could do for photographers.

How to store images on blockchain?

As described in our blockchain intro blog post, it makes a lot of sense to store ownership information for images on a blockchain. When we decided that we want to go down this road, the first important question that came up was: how can we store image data itself, in combination with the name of its author, on a blockchain?

The important thing to first note here is that blockchain saves data forever and if someone needs to store a lot of data, it can get pretty complicated. It’s easy to store the name of an author on a blockchain but in comparison to text data, images need a lot more storage. For example, contact details of a person need less than 1KB of storage while an image at low resolution has at least several 100KB and more likely up to several MBs. Every image consists of thousands of pixels. Even if there is a slight change to an image, the data of the image changes completely. Therefore, it was pretty clear to us from the beginning that if we want to store ownership information for images on a blockchain, we can not store only the image data itself. We have to find a new approach.

So, we thought of the ways Picter could use blockchain when dealing with so many images and its changes. If the images themselves can’t go onto the blockchain, we could somehow create a text that relates to the image. This usage of text would then not take up as much data on blockchain as the image itself would, that’s how the idea for Picter’s image fingerprint was born.

Transferring visual information into text

Our research pointed us to the direction of using the image.match developed by ascribe.io. This algorithm extracts visual information like the distribution of contrast in the image and transfers it into text. The visual information of every image can be stored in 648 values that have five options from -2 to 2. Like this, the algorithm allows us to create a “hash” for every image with 5 ^ 648 different possibilities. That would be

856175526956407437403395744977000325246409657079701805965861175152583887232630757138338707681915834834333044628640334557396097276080520039978091430726254445539287845587424327993702450036120052272966991072736185371172937364708536117556727959244408117307652364713731466182646836367911115292507818264540439714486513977705912717933384200219127044689186015794189981793180723481194895286769326368700255157210540890425944093866661432912223972380161285400390625

possibilities!

This hash is very powerful since it contains the visual information of an image while it needs only limited storage space.

Image: image signature with 648 values.

Picter’s image fingerprint

We decided to visualize the hash of each image in a little box that might remind you of a QR-code. We call this the image fingerprint. It is a unique and reduced code that represents the visual content of your image. The image fingerprint allows you to recognize similar images.

Similar images have similar image fingerprints. If the image fingerprint of two images match completely this means that it is the same image. If the image fingerprint differs just a bit, the images also just differ a bit. If the image fingerprint differs a lot, then it’s different images. The more the image fingerprints differ, the more their visual content differs.

When you go to Picter and access the image details of your uploaded image, you are able to see the image fingerprint for that particular image. For now, you won’t be able to compare your image fingerprints with image fingerprints of other photographers. But in the future, we might for example, be able to notify you of images that look very similar to yours and that could indicate that your copyright has been violated.

This is our first step and foundation towards blockchain. With our image fingerprint, we are now able to store image data together with information on the author on a blockchain.

What are the next steps for Picter with blockchain?

Right now we are experimenting with storing image fingerprints on an internal blockchain. This is the next step towards reaching our big goal: to create a standardized infrastructure for frictionless and secure image-delivery between producers and consumers.

In the long term, we want to save authorship information on a decentralized blockchain that is not controlled by Picter. This way we can create a stable, reliable and transparent source of truth for authorship of images. To reach this point, we will have to continue experimenting further with blockchain. We will keep you posted about our progress, read more here soon on Medium, or sign up with Picter.com and subscribe to our newsletter!

Are you interested in this topic? Let’s get in touch! Drop us a mail at hello@picter.com or comment below.

Richard Stromer,

Chief Product Officer at Picter