The visual content of your business has a life after its publication. What will happen to your photos or videos once published, emailed or downloaded? Whether it is to calculate your marketing ROI, protect your brand, your company’s reputation, your IP or copyright, it would be better for you to know.
Your enterprise may already have a digital assets management system (DAM) or a CMS, a shared library as central point of management, including rights. But how could you extend this management beyond your servers? How to maintain the link with your content and still monitor its “life outside” ?
This article gives an overview of the technologies involved to monitor visual content on the web and their technical locks. It also shows how we remove them, at IMATAG.
How to build an efficient monitoring solution in four steps:
- Step 1 : Build and feed an Image Search Database (parsing, crawling)
- Step 2 : Recognise images and videos using Visual Recognition
- Step 3 : Identify and certify matches using Watermark Detection
- Step 4 : Monitor your assets on the internet like a radar (reports and dataviz)
Step 1 : Build and feed an Image Search Database (parsing, crawling)
Your images are on the web like a needle in a haystack and finding them requires specific resources. This phase of image search starts with a crucial and decisive step for the success of your monitoring: the web-crawling, i.e. browse the web and collect as many images as possible, which is more complex than it seems.
First, crawling robots, though peaceful, can be blocked by sites that defend against cyber-attacks. Also, access to content may require permissions (paywalls, platforms and social networks). Then comes the problem of storing and indexing the database of gathered images.
Depending on your activity, you may also be required to search for your content in printed media (books, magazines, catalogs …). In this case the search service must also be able to process the PDF files to extract the visual contents.
Searching for images in videos also requires specific treatments.
Finally, it may be necessary to target the activity of crawling robots on particular sites, identified upstream (sites of your distributors, your competitors, your customers …).
Imatag’s Smart Crawler
IMATAG’s efficiency in crawling relies on its ability to accumulate sources of web images (from partners and its own crawlers) and to process PDFs containing images. We are also experts in metadata (IPTC, EXIF, XMP) and use the results of our “search by credit” engine to parse sites where your credit is most used. Specific partnerships can be set up for access to certain images despite the presence of paywalls or bot-blockers.
Learn more about the difference between parsing and crawling.
Step 2 : Recognise images and videos using Visual Recognition
Crawling allows you to collect images, then you have to index them and then store them in base (we will not detail this part). Then there are two options: recognize them visually, or read their metadata.
Reading metadata is much easier, because any basic system can read a string and find your brand. But in fact, 97% of the images on the web are stripped off their metadata (as shown in our study ).
So let’s go for the first option : visually recognize your images. For this visual recognition will be useful but may not be sufficient. Indeed the similarity between some images can be misleading, or on the contrary undetected after too important modifications like cropping, changes of colors, symmetry …
The trick: use Digital Watermarking to improve visual recognition or to distinguish apparently identical copies (non exclusive photos, transactional watermark, tracing leaks).
IMATAG’s invisible watermark
To overcome the uncertainty of visual recognition, IMATAG uses patented Digital Watermarking technology. It is a way to imperceptibly insert information into the pixels of the image, such as a unique identifier for example. The watermark acts like an invisible QRCode hidden in the pixels.
This method is much more robust than a visible watermark (easily erasable), and allows to certify the origin of an image when this invisible mark is detected by the recognition algorithm. The link with the original metadata is then maintained.
Despite a compression, a color change, a crop, the invisible identifier is found and makes the link with the original image (see demonstration below).
Step 3 : Identify and certify matches using Watermark Detection
Applying a Digital Watermark on all your visual assets allows to distinguish seemingly identical copies (non-exclusive photos, transactional watermark, leak tracing).
Not only this technology will allow you to find your images with better results than simple visual recognition, but it will also allow you to obtain distinct information from multiple copies of a single image.
For example find the ID of the user who downloaded this copy from your DAM. This information can be very useful in case of premature leakage of embargoed photos, for example.
This feature is called the transactional watermark or traitor tracing.
Step 4 : Monitor your assets on the internet like a radar (reports and dataviz)
A long list of images corresponding to yours found on the web is a source of information, but it will be much more exploitable presented in consolidated and visual form.
An alert service is also useful to act quickly and to avoid a daily consultation of the reporting interface.
Data about the sites on which your images were found is a plus for your analysis: audience, monetization, category …
IMATAG’s monitoring dashboard
Notifications issued each time you detect images allow you to quickly analyze whether their use is authorized or compliant.
An online report gives you a statistical view of the sites where your images are used.
Additional information such as the world rank of these sites (audience indicators) helps you assess the exposure of your content and make the right decisions.
Now, who is using your visual content?
IMATAG Monitor has been developped on op of the 4 skills listed above. You can use it online via our web service, or integrate it into your system using our API.
Learn more about IMATAG: