A markov-renewal-modulated tes model for jpeg-like compressed "star wars" video

The Bulk Of Previous Work In Video Source Modeling Has Been Confined To Short Se Quences Of Empirical Records, Which Attempt To Capture The Autocorrelations Of Empirical Bit Rate Records. This Talk Presents A Novel Approach To Video Modeling With Two Main Contributions. First, It Presents A Model Of A Full-length Movie (a Jpeg-like Version Of "star-wars"); And Second, It Captures Both The Distribution And Autocorrelation Functions Of The Empirical Record. The Modeling Methodology Developed Consists Of Several Stages. The Video Se Quence Is First Segmented Into Scenes. Individual Scenes Are Then Classified (by Mean Bit Rate) Into Classes, Each Of Which Is Assumed To Consist Of Sample Realizations From The Same Probability Law - In This Case A Tes (transform-expand-sample) Process. The Classification Is Accomplished Via An Iterative Heuristic Procedure For Partitional Clustering Of The Set Of Scenes, Having Identified A Suitable Set Of Partitioning Criteria And Parameters. Next, A Tes Model Is Derived For Each Class; Each Tes Model Simultaneously Captures The Empirical Marginal Distribution (histogram) And Autocorrelation Function Of The Corresponding Scene Class. Scene Transitions In The Movie Are Modeled As A Sim Ple Markov Chain Whose State Space Consists Of Just Four States (scene Classes). Scene Durations Are Modeled As Class-dependont Renewal Processes, One For Each Class. Thus, The Entire Movie Is Modeled As A Markov-renewal-modulated Tes (mrmt) Process With Class-dependent Holding Times, Generalizing The Well-known Markov-modulated Poisson Process (mmpp). Having Generated A "synthetic" Monte Carlo Simulation Of The Bit Rate Process Of "star Wars", The Simulation Statistics Were Compared To Their Empirical Counterparts And Were Found To Be In General Good Agreement. The Results Support The Efficacy Of The Modeling Methodology And Suggest That A Benchmark Of Realistic And Compact Source Models Of Full-motion Video, Suitable For Driving Realistic Monte Carlo Simulations Of Broadband Telecommunications Networks, Can Be Developed By This Methodology For Other Types Of Movies Under A Variety Of Coding Schemes.