Content based video identification

Abstract

<P>IN THIS THESIS WE FOCUS ON CONTENT BASED VIDEO IDENTIFICATION PROBLEM. WE REQUIRE IDENTIFICATION BE ROBUST TO SHOT EDITING, VIDEO FORMAT CHANGE AND COLOR DISTORTION. WE PROPOSE TWO METHODS TO SOLVE THIS IDENTIFICATION PROBLEM: ONE IS SEQUENCE MATCHING; THE OTHER IS PATTERN SET MATCHING. IN SEQUENCE MATCHING CATEGORY WE USE COMPOSITE HIDDEN MARKOV MODELS TO REPRESENT AND RECOGNIZE SHORT VIDEO CLIPS. THIS APPROACH CAN EFFICIENTLY RECOGNIZE VIDEO CLIP WITH SHOT REMOVAL. IN PATTERN SET MATCHING CATEGORY WE PROPOSE A NOVEL COLOR FEATURE REPRESENTATION FOR SHOT. THIS REPRESENTATION IS NOT ONLY ROBUST TO IMAGE QUALITY REDUCTION, FRAME SIZE AND FRAME RATE CHANGE, BUT ALSO TO COLOR DISTORTION SUCH AS BRIGHTNESS/CONTRAST ADJUSTMENT. THE VIDEO SEQUENCE IS MODELED BY A SET OF SHOT COLOR FINGERPRINTS, AND A VIDEO SIMILARITY MEASURE IS DEFINED TO IDENTIFY SHOT CHUNKS. WE CONDUCTED EXPERIMENT ON 100 VIDEO CLIPS, AND QUITE LOW IDENTIFICATION ERRORS ARE ACHIEVED.</P>