Image/Video Communication And Storage Towards Cloud Mobile Media Computing

Growing popularity of mobile devices,together with the development of wireless Internet,greatly enriches the user experience on multimedia.However,the limited on board resources and the time varying wireless channel severely constraints its further development.Fortunately,the rich computing resources in cloud computing naturally offer a solution to reduce the cost on mobile device,which results in a new research area—cloud mobile media computing.It brings both opportunities and challenges for traditional image/video communication and storage.First,with the flourish of multimedia on the web,it is easy to find similar images for a query,especially landmark images.Traditional image coding such as JPEG cannot exploit correlations with external images.Existing vision-based approaches by reconstructing from local descriptors are able to exploit such correlations but cannot ensure the pixel-level fidelity of the reconstruction.Second,the inherently time-varying and unreliable wireless channel limits the communication bandwidth between mobile and cloud.The current researches cannot achieve both channel signal-to-noise ratio(SNR)and bandwidth scalability in a universal scheme.Besides,they cannot exploit the external correlation in the cloud to improve the transmission efficiency.Finally,redundancy is necessary for a storage system to achieve reliability.The frequent errors in a large-scale storage system,e.g.cloud,make it desired to reduce the cost to recover from errors.Among all types of data in the cloud storage,videos have occupied a significant part due to the high volumes and the rapid development of video sharing and video-on-demand services.Most redundancy protection solutions treat all files as general data,in which any unrecoverable bit error will lead to permanent loss of the whole file.Unlike general data,videos can tolerate certain level of quality degradation.To solve the above problems,this paper focuses on how to improve the image coding efficiency via exploiting the external correlation in the cloud while ensure low coding complexity and high pixel-level fidelity;how to achieve both SNR and bandwidth scalability and exploit the external similarity with images in the cloud in a universal scheme;and how to achieve a better trade-off among reliability,storage cost and reconstruction cost via exploiting the unique characteristic of videos such as scalable representation in cloud storage.The main contributions of this thesis are summarized as the following three parts:In the first part of our work,we focus on photo sharing or uploading in cloud mobile media computing.We try to exploit external correlations in the cloud in terms of low coding complexity and high pixel-level fidelity and propose a cloud-based distributed image coding framework.For each input image,a thumbnail is transmitted to retrieve correlated images and reconstruct it in the cloud by geometrical and illumination registrations.Such a reconstruction serves as the side information(SI)in the Cloud-DIC.The image is then compressed by a transform-domain syndrome coding to correct the disparity between the original image and the SI.Once a bitplane is received in the cloud,an iterative refinement process is performed between the final reconstruction and the SI.Moreover,a joint encoder/decoder mode decision at block,frequency and bitplane levels is proposed to adapt to different correlations.Experiment results on a landmark image database show that the Cloud-DIC can largely enhance the coding efficiency both subjectively and objectively with up to 5d B gains and 70%bits saving over JPEG with arithmetic coding and perform comparably at low bit rates with HEVC intra coding with a much lower encoder complexity.In the second part of our work,we consider efficient image transmission via time-varying channels in cloud mobile media computing and propose a new distributed compressive sensing scheme for wireless image transmission that can leverage similar images in the cloud.It is featured by exploiting the external correlations in the cloud and achieving channel SNR and bandwidth scalability,high efficiency and low encoding complexity in a universal scheme.For each image,a compressed thumbnail is first transmitted after forward error correction(FEC)and modulation to retrieve similar images and generate a side information(SI)in the cloud.The residual image after subtracting the decompressed thumbnail is then coded and transmitted by compressive sensing(CS)through a very dense constellation without FEC.The linearly and ratelessly generated CS measurements make it capable of achieving both graceful quality degradation(GD)with the channel SNR and bandwidth scalability in a universal scheme.Before that,a mode decision and transform-domain power allocation is introduced for better bandwidth usage and protection against channel errors.At the decoder,a two-step distributed compressive sensing decoding is performed to recover the residual signal,where both the local and nonlocal correlations within the image and that with the SI are exploited.Simulations on landmark images and a AWGN channel show that the received image quality gracefully varies with the channel SNR and bandwidth.Furthermore,it outperforms existing schemes both subjectively and objectively by up to 11 d B gains compared with the state-of-the-art transmission scheme with GD,i.e.Soft Cast.In the third part of our work,we focus on the basic video storage problem in cloud storage.We investigate the multi-layer video representation such as scalable videos and simulcast streaming and propose an unequal error protection scheme for scalable video storage in cloud.By taking the local reconstruction codes(LRC)as an example,which has been employed in Windows Azure Storage,and providing less protection for less important layers or video copies,a better trade-off between storage and the repair cost could be achieved.Both theoretical and simulation results show that such a better trade-off can be achieved over the LRC with equal error protection,though the recovered video quality might be slightly degraded in a rare case.