Research And Implementation Of Picture In Picture Algorithm Based On Video Bitstreams

Picture in Picture(PIP) technology which can display different image in different position of the screen, was originally used in the analog television system. However, with the constant development of digital video compression technology as well as the popularity of the internet, PIP display technology based on video bitstreams is capable of combining multiple streams into one, therefore, the information storage space and transmission bandwith are reduced, as a result, this technology has attracted attention in recent years. In this thesis, PIP display algorithms is studied based on video bitstreams in H.264video coding standard, and a spatial resolution transcoder based on "information re-used" is proposed in pixel domain. The main work of this thesis is presented as follows:(1) The application background of PIP display is simply introduced, and the research significance along with current research status of PIP display are summarized.(2) A fast decision algorithm of prediction type is presented, which can obtain prediction type in encoder quickly according to that in decoder under certain judgment criterion.(3) A fast intra mode decision algorithm is proposed based on statistical information. In this algorithm, the prediction mode extracted from the decoder is used to synthetic or map that in the encoder, therefore, the amount of calculation in the intra mode research is reduced.(4) In H.264standard, macroblock using inter prediction requires segmentation search process. However, in this thesis, the search process is no longer used during encoding, and a fast decision algorithm is presented instead. According to the high correction before and after sampling, the edge parameter of the macroblock in the decoder is calculated to determine the segmentation mode of the corresponding macroblock in the encoder. After that, for the macroblock having achieved segmentation mode selection, motion vector composing and refinement is completed.The algorithms proposed in this thesis are implemented under VS2010intergrated development, through the analysis and comparision of the experimental results, the performance of the algorithms are verified.