Research On Decision Taking&Optimal Algorithm In Content Based Adaptive Transmission Of Soccer Video

Along with the advance of multimedia technologies and network communicaitons inrecent years, as well as the increasing ease of generation, processing and access of videodata, more and more users prefer accessing diversity multimedia content with the mobileterminals and handheld devices. The current multimedia services are limited by thecomplexity of usage environments, such as the heterogeneity of network characteristics,the diversity of terminal devices and diversity user preferences, which urges thedevelopment of universal media access (UMA). Content based adaptive transmission(CBAT) of video stream utlitlizes a lot of content characteristics of original videos, to rankvideo contents with semantic features to user’s requirements, makes an optimal decisionon manipulating the transmitted video clips to adapt the constraints of heterogeneouty ofusage environments, and achieves higher quality for those important contents than othersless important. The challenge of UMA has lead to hot research and development on CBAT,which are involved in several key issues, such as semantic content description andclassification, adaptive transmission decision-taking model, adaptation engine and qualityevaluation. In order to solve the issues on content-based adaptive transmission decisionand optimal algorithm, soccer video is taken as research object to study CBAT in deepth.For the optimization on subjective quality of user’s prefered content in adaptive videotransmission, an archichecture is proposed on semantic content based on adaptive videotransmission. And then, temporal and spatial descriptions of semantic contents are defined,the importance of semantic content is ranked by user’s preference interactively, amulti-objective utility optimizing model is set for adaptation decision on semanticcontents according to user’s preferences, the content based adapation is defined asselection/dropping on different granularities of sematic units in temporal-spatial domain,the perfromance of adaption is evaluated by means of objective quality and subjectivequality.According to the syntactic structure and adaptation objective for temporal-spatialsemantic contents, an optimal model on adaptation decision based on temporal-spatialsemantic contents is proposed. Content of soccer video can be decomposed into semanticunits with strict syntactic structures in temporal-spatatial domain respectively, theadaptation decision is settled on each semantic to select/drop different ranularities of content to maxmize subjective utility under a certain constraint, content based adapationdecision is separately set as two optimal models for the different application. Among them,the temporal adaptation decision is modeled as Multidimensional Knapsack Problemsbased on different grannulaties of semantic units, and the spatatial adaptation decision ismodeled as optimization model for the semantic object based regions of interest in screen.In order to solve the adaptation on bandwidth constraint for Internet real-timestreaming on-demand services, a real-time adaptive transmission based on frame droppingof semantic content is implemented. The important level, shot classification and frametype is ranked on temperol content, the ratio to drop frames on different semantic unitswhich includes event, shot and GoP is determined on condersideration of networkcongestion status, the optimal quality of transmitted video is solved in low and variable bitrate network by frames dropping on different granularities of semantic contents. For theimportant content is retained to delivery, the user’s overall satisfaction is improved mostthan other traditional means such as highlight or motion activity based frme dropping,which is indicated by the subjective evaluations.In order to solve the adaptation on heterogeneous display performance for diversityterminal equipments, a video retargeting method based on user interested object for soccervideo is implemented. Optimal region of interest in different types of shots is determinedon the coordinates of user interested object, and then content in the region of interest iscropped and scaled to terminal’s screen as its resolution and aspect ratio. Not onlytransmits adaptive video content under the limitation on different resolutions and aspectratios of heterogeneous terminals, the proposed method also indicates great improvementon visual quality and semantic quality.Many further research work are proposed to content-based adaptive video transmissionsystem of soccer video, intellectual learning to rank semantic content based on user’spreference should be improved, the corresponding adaptation decision taking engine andadaptation engine look forward to be applied in more adaptive engine Internet videostreaming technologies, taking real-time multi-dimensional adaptation decision anddesigning an end to end cross-layer application based on MPEG-21standard will also be afocus in future work.