An Unsupervised Approach to Video Shot Boundary Detection Using Fuzzy Membership Correlation Measure

In this paper we propose an improved cut detection or shot detection algorithm adapted to any domain of movies with various result sets. Shot is actually the series of interrelated consecutive pictures or frames taken from a film or part of a film contiguously and representing a continuous action in time and space. Consecutive two different shots produce an important visual discontinuity in the video stream which is called a cut. Here the video shots are assumed to be fuzzy sets and the fuzzy correlation between them is defined on the same universal support. It is shown that Spearman´s rank correlation coefficient can be applied if the members of the supports are ranked according to the fuzzy membership values of each set. Next a membership-value-based fuzzy correlation measure is explained with the experimental result. Results indicate encouraging avenues for detection of hard cuts with high precision.