Detection of hard cuts and gradual transitions from video using fuzzy logic

Abstract

In this paper, we introduce a novel approach for the detection and classification of shot boundaries based on interval type-2 fuzzy logic. We consider fuzzy low level features (Fuzzy Histogram (FH) and Fuzzy Co-occurrence Matrix (FCM)) and evaluate a fuzzy rule base modelled by interval type-2 sets to detect cut boundaries. After detection of cuts, a recovery mechanism is used to reduce false positives caused by illumination variation (e.g., camera flash light). The fuzzy rules used for gradual transition form a unified approach to detect various types gradual transitions like dissolves, fades and wipes, by a single method. Test results indicate that very high precision and recall could be achieved for the test videos, particularly the recall value of gradual transitions.