2.3.5 Results – Video Structure Meaning

• size,
location on screen,
position relative to other objects,
duration on screen, and
movement on screen.
Points of signicant change are noted; the degree of change considered signicant is user selectable.
As an early experiment to test such an approach to analysis, we attempted to discriminate between
two moving image documents on the same topic—women running the marathon. In several infor-
mal screenings of the two documents, we gathered viewer opinions of the two videos. Document
A was almost always described as dynamic, exciting, engaging, emotional; Document B was almost
always described as dull, boring, snoozer. Yet, both documents showed approximately 50 women
running 26 miles in urban settings.
Each and every frame (1/30th sec-
ond) of the two marathon documents was
observed on the computer monitor and
measurements were entered directly into
the program. e software calculated values
for several variables at each frame and then
made comparisons across frames. is frame
by frame analysis of the two marathon
documents does make apparent signicant
dierences in structure, and gives a precise representation of what might cause the term dull to be
applied to one document and not the other.
e test system marked each point in each document where the area occupied by the pri-
mary object changed by 15% or more; and each change in object (e.g., head and shoulders image
of Joan Benoit in one frame, then body of Grete Weitz in the next frame). e frames between
any marked point of change and the succeeding marked point of change are termed an image set.
e threshold for % object area change is variable from 0–100% within the test environment. e
discrimination test used 15% as a small but signicant number for the pattern recognition capabil-
ities of the human visual system. A gure much lower than this is easily discerned, but it will likely
present changes which result merely from minor (dierence that doesnt make a dierence) motions
by the objects or from camera motion unavoidable in documentary recording. For example, in the