A series of memorable and forgettable photos from the study (Credit: Oliva and Torralba labs , MIT)

According to Phillip Isola, MIT graduate student who took part in the research:  “Pleasantness and memorability are not the same”. Despite initial concerns the MIT team found remarkable consistency among hundreds of people who participated in the memory experiments conducted as part of the research.

One of the direct outcomes of this research was the development of a a unique algorithm that can rank images based on memorability. According to the MIT researchers this algorithm might prove useful in the future for graphic designers, photo editors and others who might want other people to remember their images.

Previus research done by one of the MIT team members found out that the human brain can remember thousands of images, with a surprising level of detail. However it was also found that not all images are equally memorable.

The current study combined a collection of about 10,000 images of all kinds and let test subjects look at a series of them (some of which were repeated) and then press a key on their keyboard, when an image appeared that they had already seen. The image "memorability rating" was calculated by how many participants correctly remembered seeing it.

According to another member of the MIT research team: “There are always differences between observers, but on average, there is very high consistency”. The last part of the test included a different task - each participant in the test was asked to label all the objects in all the images he observed. Later on a computer model analyzed those marked images to determine which objects make an image memorable.

 

The results of the test showed that in general, images with people in them are the most memorable, followed by images of human-scale objects and interestingly also close-ups of objects. Least memorable on the test were landscapes, although those can be memorable if they feature an unexpected element, such as a strange looking tree with a familiar shape.


The last stage of the research included the development of a computational model that can analyze the images and their memorability as rated by humans. For each image, the computational model analyzed various statistics — such as color, or the distribution of edges — and correlated them with the image’s memorability. Using this model the MIT researchers were able to create an algorithm that can predict memorability of  new images - ones that the computational model hasn’t encountered before.

 

Future applications of the algorithm can also include developing mobile app for an iPhone or similar device which can rate the level of  memorability an image taken by the device has. A similar technology could in the future be incorporated into cameras.

 

TFOT has already covered several memory related studies including "memory encoding methods" - a research conducted by scientists from the University of California at Santa Barbara which uncovered a process used to encode memories in the synapse layer connecting neurons of the human brain and "IBM 'cat brain' project ".

More information on the MIT memorable image research can be found on their original paper (PDF).