How to clean up images to improve search results in GEOINT investigations

This article provides a step-by-step guide on how to clean up images to improve search results, as well as techniques and tools to help you achieve more accurate image search results. Learn how to crop and edit images, remove excess detail, adjust contrast and brightness to improve visual quality.

In a quiz, Tilman Wagner shared this photo and offered some interesting questions for analysis:

1. Who published it and where?

2. What was the occasion?

3. Was Tillman involved, and if so, how?

So, where should you start? First you need to determine the shooting location to answer Tillman’s questions. There are a few clues in the image that could be a good starting point: the text on the screen is written in British English (where the word “organization” is spelled with an “s”), hinting at a likely networking event or meeting. There is a WiFi sign, but the network name is too small to read, otherwise you could use the Wigle service for a deeper analysis. It does not give quick results yet.

One of the cardinal rules in geolocation (and research in general) is that the more unique or distinctive an object is, the more useful it is for identifying where it was taken. Undoubtedly, the most noticeable feature is the unusual pyramid-shaped building that can be seen from the window. If it can be found, it will be possible to determine the shooting location.

Cropping the image and using reverse search seems like the most obvious way to find the building, but in this case, this approach has a significant drawback. When people look out of a window, they instinctively ignore details such as frames or curtains to focus on objects outside. Unfortunately, reverse image search algorithms are not yet capable of this: they do not distinguish between foreground and background elements that matter. So when you try to solve this problem in this way, the result is likely to be a series of matches that look something like this:

Reverse image search engines show their limitations here. They focus on the window frame and curtains, but not on the outside of the building that interests us. So what can be done?

Removing noise from an image

In order to perform a reverse image search more effectively, it is necessary to simplify it by removing unnecessary foreground elements. The Cleanup.pictures service is an ideal solution for this – it is easy to use and allows you to easily remove unwanted details, for example, window frames. Instead of leaving empty areas, the tool automatically fills in the removed areas based on the rest of the image and guesses what details should be displayed. This is perfect for our purposes.

The first step is to upload the cropped image to a website for further processing.

Next, we use the eraser tool to select the details that we want to remove.

Here is the end result.

The Cleanup.pictures tool uses a “smart eraser” to remove unwanted image details and fill in empty spaces, focusing on unchanged parts. Thanks to this, a “cleaned” image is obtained, which significantly increases the chances of success in a reverse search. Below are the results of reverse image searches with the foreground removed, which helps identify objects in the background more accurately.

After removing the window frame, it was possible to obtain much more accurate matches with the building in the background. It turned out to be the National Library of Latvia in Riga. Looking at the area on Google Maps, you can see that the only building that this photo could have been taken from is the Wellton Riverside Spa Hotel. The other buildings opposite the library do not have glass balconies as in the original image, confirming the accuracy of this location.

As confirmation, a quick look at the conference rooms on the hotel’s website reveals a room that matches the one in Tillman’s photo. The National Library can even be seen from the windows.

So the place is found – how to find out about the conference?

Event search

One of the geotagged images links to this image posted on October 25th. It links to the hashtag #MILSummit2022 – so we’ll have to find out what exactly that summit was.

The user also posted another photo that appears to have been taken at the same event. Judging by the name of the conference – Baltic MIL Summit – and the main topics discussed at the event, it can be concluded that this is a summit dedicated to media literacy, combating disinformation and fact-checking. Since Tilman works in this area, it confirms that the direction of the research is correct and related to the event where similar topics were discussed.

Researching the MIL Baltic Summit quickly revealed that the event was held at the Wellton Riverside Spa on Monday 24th October and was hosted by Žinių Ekonomikos Forumas. The event’s Facebook page has hundreds of photos of conference attendees, which not only confirms the location of the event, but also provides additional evidence—among the photos, we even found the same image that Tillman used in his test. This is a perfect confirmation of the correctness of the chosen path!

Photos from the event hint at the close collaboration between DW Akademie and the MIL Baltic Summit held in Riga, an area in which Tilman works. This increases the suspicion that Tillman may have had a connection with this conference. At the same time, for confirmation, it is important to refer to the details: on the website of the organizer, Žinių Ekonomikos Forumas, there is a publication dedicated to the summit, as well as a schedule of events, which lists the speakers and their topics, is available.

Representatives of the DW Akademie do appear among the speakers and participants, but Tillman is not among them. This is an important fact because, although DW Akademie was directly involved in the conference, Tillman himself is not listed in the lists of speakers or participants. Also, looking through the many photos on the event’s Facebook page, you can see that he was not captured in any of the pictures, even though many of the other participants were photographed in detail.

The absence of his name from the official schedule and photos from the event probably means he did not attend the summit in person. This may be an indication that although his organization participated in the summit, his role was remote or organizational, with no physical presence in Riga.

Some of Tillman’s colleagues spoke at the event, and while there is no direct evidence that he attended in person, the schedule indicates that certain sessions were streamed online, so it is possible that Tillman could have joined them remotely. However, this is difficult to confirm without additional sources.

Instead, another approach can be considered: check how DW Akademie covered the summit on its news platforms. One-way communication via Žinių Ekonomikos Forumas (ZEF) already shows DW Akademie’s participation through conference photos and schedule, but further investigation of DW Akademie’s own channels may help clarify the details.

The DW Akademie website has two social media feeds: X and Facebook. Posts from October 24-25 show that the conference was available for viewing online, supporting the possibility of Tillman’s remote participation. In addition, significant attention was paid to the presentation of Lena Riche, who presented the Go Verify game as part of the summit program, as evidenced by active advertising on Facebook.

These facts suggest that, although there is no direct evidence of Tillman’s presence, his organization was actively involved in the event, and he could probably have been involved remotely, given his professional interest in the summit’s topics.

And here on X…

The X message contains a direct link to the Go Verify game, which is the main theme of Deutsche Welle’s advertising around the MIL Baltic Summit. After following the link and reading about Go Verify, we find a familiar name at the bottom of the article: