Computational Product Presentation Video Editing Framework Based on Editing Attribute Constraints

Nowadays,there is an obvious increasing demand for product presentation videos(PPVs)in the field of e-commerce.In order to produce high-quality PPVs,the importance of video editing has been emphasized.When comparing with video editing for advertising,PPVs pay more attention to the product information.Considering the effect of information transfer,we conducted two experiments to explore the factors(information arrangement,editing coherence,editing rhythm and mise-en-scène rhythm)that affect consumers’information perception.We applied those elements to build an automatic editing framework based on editing attribute constraints.We at first divided footages into clips under the constraints of information type,mise-en-scène rhythm and duration.Then we use the hidden Markov model and the Viterbi algorithm with global constraints to realize clips editing.Finally,we developed an automatic editing system prototype for the garments.Compared with the professional hand-edited result,our method can complete video editing more efficiently,and our outputs are quite close to the hand-edited result.