Fraktalna i multifraktalna karakterizacija 3D video formata

Abstract

Trodimenzionalni (3D) video sadržaj postaje sve prisutniji u saobracaju žicanih i bežicnih mreža. Kako 3D video karakteriše velika kolicina podataka, njegovo prisustvo u mreži predstavlja znacajano opterecenje za saobracaj mreže, pa je neophodno poboljšanje u nacinu prenosa ove vrste signala. Predmet ovog rada je fraktalna i multifraktalna karakterizacija 3D video formata sa ciljevima pokazivanja da 3D video ima fraktalne i multifraktalne osobine i opisivanja ovih svojstava, pore enja svojstava za razlicite video formate i mogucih primena u usavršavanju modelovanja saobracaja 3D video formata i razvijanju boljih tehnika za prenos, smoothing i statisticko multipleksiranje videa. Izvršeno je kreiranje algoritama i softverska realizacija u programima Matlab i Python, u okviru matematicke metode iz oblasti kompleksnih sistema i haosa, za odre ivanje fraktalnih osobina (metoda agregacije varijanse (aggregated variance method), R/S statisticki metod (R/S statistic method) i metod više skala (multiscale method) za odre ivanje Hurst-vog indeksa) i multifraktalnih osobina (metoda momenta za multifraktalne spektre i procenu generalizovane dimenzije, metoda histograma za multifraktalne spektre i inverznu multifraktalnu analizu). Na osnovu dobijenih rezultata potvr ene su fraktalne i multifraktalne osobine 3D video formata i izvršeno njihovo pore enje, uzimajuci u obzir format 3D videa, strukturu, tipove frejmova, kvantizacione parametre kodera, agregaciju frejmova i metode emitovanja...

Three-dimensional (3D) videos are becoming increasingly present in wireland and wireless networks, and due to the high data content they present significant load in the network. This impact of 3D video on network traffic causes a need of improving means to handle this type of data. Subject of this thesis is fractal and multifractal characterization of 3D video formats with the main goals to show that 3D video has fractal and multifractal properties, to describe them, to compare these properties for different 3D formats, for possible applications in traffic modeling of 3D video formats and for the development of improved transport techniques for smoothing and statistical multiplexing of video. Algorithms and software realizations for mathematical methods in the area of complex systems and chaos were developed in Matlab and Python, for determination of fractal properties (the aggregated variance method, the R/S statistic method, and the multiscale method) and multifractal properties (the method of moments for multifractal spectra and estimation of the generalized dimensions and the histogram method for multifractal spectra and inverse multifractal analysis). Based on the obtained results, fractal and multifractal properties of 3D video formats are confirmed, their comparison is performed, for various 3D video formats, video structures, frame types, values of the encoder quantization parameters, aggregation of the frames, and streaming method of video...