基于结构化学习的数据压缩方法.pptx

VALSE-Webinar LectureStructured Modeling and Learning in Generalized Data Compression and ProcessingHongkai Xiong熊红凯电子工程系上海交通大学13 Jan. 2016LSparse Representation=N Sparse representation where xΨθΣMultimedia CommunicationHigh dimension3D signal（video）1D signal（audio）2D signal（image）SVCSourcesDVCVideo coding: advances in higher dimension and higher resolution, in goals of better R-D behavior and greater compression rateNetworks: develops to multiple data streaming within heterogeneous network structure, in goals of higher throughput and transmission reliabilityMulticastMany-to-manyMulticastOne-to-manyUnicastOne-to-oneNetworks Generalized Context Modelingin Signal Processing Wenrui Dai, Hongkai Xiong, J. Wang, S. Cheng, Y. F. Zheng, Generalized Context Modeling with Multi-Directional Structuring and MDL-based Model Selection for Heterogeneous Data Compression, IEEE Trans. Signal Processing, 2015.Wenrui Dai, Hongkai Xiong, J. Wang, et al., “Discriminative structured set prediction modeling with max-margin Markov network for lossless image coding,” IEEE Transactions on Image Processing, 2014.Wenrui Dai, Hongkai Xiong, X. Jiang, et al., “Structured set intra prediction with discriminative learning in max-margin Markov network for High Efficiency Video Coding,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 23, no. 11, pp. 1941-1956, 2013. Heterogeneous Data Compression: DataHeterogeneous data are generated by multiple interlacing sources complied with different incomplete distribution statisticsImage Video: Spatial correlations are characterized by piecewise smooth with local oscillatory patterns like multiscale edges and texturesGenome sequence: Repeatable patterns of nucleotides in various regionsExecutable files: Multiple interlacing data streams, e.g. opcode, displacement, and immediate dataHeterogeneous Data Compression: FrameworkData to be predictedEstimated probabilityEncoded bitstreamContext ModelCode…Structured probabilistic model