Multimodal Semantics Extraction from User-Generated Videos
Algorithm 3
Selection of optimal cameras.
(1) Determine the main point of interest P, by applying the method described in Section 3.3.
(2) Consider all the lines which intersect the main point of interest, each line having a different slope m and representing a candidate separating line. For each line, count the number of cameras that lie on each of the two sides of the line.
(3) Select the candidate separating line which yields the maximum number of cameras on one of its two sides. The selected line represents the optimal separating line, and the optimal set of cameras to be used for generating video mash-ups is made of those cameras on the most populated side of the optimal separating line.