A straight line-based distance measure to compute photographic compositional dissimilarity

초록

According to a photography theory, lines are an important element that creates the composition and mood of a photo. In this paper, we propose a measure to compute photographic compositional dissimilarity using lines. To develop the distance measure, we investigate some attributes of line that classify composition of photos for line-used photos. And we implemented an image searching system which retrieves photo compositionally similar to given query to evaluate performance of the proposed measure. The searching system shows the precision of about 90% maximally and was capable of reliably retrieving compositionally similar to given query even if some objects were included in photos. © Springer-Verlag 2009.