Interactive Logical Story Unit Segmenta- tion

Logical Story Unit (LSU) segmentation in general domains, such as movies and television series, requires interaction between human experts and automatic tools. We present an interaction model that allows users, who are non-experts in the field of video processing, to segment videos into logical story units by tuning models rather than manual correction of results of automatic methods. Suitable features are determined interactively by visualizing their values in the context of the original video data. The VidDex system developed uses backprojection on various video visualization modes to meet presented requirements for good visualization. Evaluation on two Hollywood movies shows that systematic adaptation of LSU-segmentation parameters is more effective than adaptation of the segmentation results. More importantly it allows for storage of human judgement on computed similarity functions.