Media Visualization – changing perspective
This essay is written for Dynamic Visualization course / Alto Mlab.
Million details
Imagine standing on a busy city street. In a glimpse, your eyes capture the landscape and you brain analyses thousands of various details that fall instantly into combination of patterns. You see pedestrians, public transport, trees, garbage, hipster guys with mustaches toggling their iPhones, giggling girls, pooping pidgins and so one.
Despite this enormous amount of data, one will not get confused simply because at this stage the primary object is to grasp an overall picture – having an light observation of a landscape and it’s content. Anxioty might occure if one is already looking for something specific and especially if limited by time or other resource; for instants running late from work and desperately looking for a taxi. In this case, the person is subjected to a keyword and is ‘browsing’ the landscape in front of oneself – looking for anything that has a shape of a car, is close, is yellow and has a bright light on the top of the cabin. But when there is no such goal, the question that drives one’s observation is not “where is it” but “what’s here in general?”.
Manovich (2011) points out that latter question is quite rare in media related research. It is obvious that due the enormous amount of media data available , it is outside our abilities to actually see the content of them all.
From libraries, tv channels and movies to web classification manners such as grids, lists, slides, galleries and thumbnails the main object is to show a capture of larger collection. What these methods do not show is an overall shape of the given collection. One can surely utilize metadata, but it is still not fully adopted; even the best examples of it do not reflect on the content itself in a same rich manner as our human comprehension does when we actually see the content. Not to mention the data access itself: search machines are based on the idea that person knows what she is looking for, excluding Google’s beeing lucky -randomizer. (Manovich, 2011.)
In other words, data set of hundreds items with metadata notion “tropical island videos” gives far less information regarding content patterns as does original conclusions drawn by a person who has actually seen videos of different tropical islands.
Dealing with visual data
Eleven Montage
In his paper Manovich (2011) presents the idea of media visualization – a methodology to deal with large collections of videos and still images. Usually media is studied through translation into numbers and dealing with differences between the numeric results. In media visualization the outcome is another media, thus the result is reposition of media that was at hand.
Manovich (2011) proposes three main techniques.
Image montage collects images or video frames in one larger overview or in other words, zooming out. Images can be brought together and organized into tight image grids by using even the most basic metadata such as production date, size or creator. This allows to see new patterns, locate items that stand out, detect clusters and spot differences.
Temporal and spatial sampling selects a subset of images from image sequence, such as page numbers, frames and upload dates or image parts, such as cropped area.
Remapping rearranges the data accordingly to a specific variable, such as color saturation or amount of movement in one video shot.
Reflections
Manovich turns around the ground idea of media consumption. We are inevitably used to deal with single media items at time – one movie, one book, one picture. It is clear that this should be the case in most situation – we can only concentrate on one to few ‘reality mapping’ at the time. But it is also true that when zoomed away enough, we can also deal rather easily with enormous amount of data. This also changes the nature and results of our observation: we begin to spot more general differences and similarities, see re-occurring patterns and can summarize more delicate outlines of the data set in question.
I see Manovich’s note more enlightening and groundbreaking than the actual visualization techniques that were used for creating examples. His approach turns current search-driven media research upside down, opening great possibilities for practical implementation.
