Shot classification for hard transition

Long transition analysis is a post-shot boundary detection technique used to process special effects once they are detected. Shot classification is one of the stages in such a process, used to classify pixels within the hard transitions, such as wipe, mosaic and push, once detected. In this paper, we present the fundamental algorithm and improvement for the technique. Such classification can achieve an accuracy of over 95% for low disturbance sequences, and on average, over 90% for most cases.