Article Preview
TopIntroduction
The increasing demand of electronic gadgets in multimedia applications expects high transmission efficiency and better video quality. Various coding standards have been evolved for the past two decades to improve transmission efficiency, and efforts have been made for optimizing the architectures of various error concealments (EC) used in video codec for improving video quality. The traditional EC algorithms are dependent on large block size, significant edge details, iterative process, and hence, results in computational complexity. Also, due to fluctuation in bandwidth, particular EC cannot retrieve entire lost macro-blocks (MBs). An effort has been made to develop an adaptive EC predictor that can recover more significant features or structures of entire lost MBs in video-codec compression schemes. Being open source, the work is focused on baseline profile of AVC/H.264 coded and later the work is generalized for recent video codec. Most of the research on temporal error concealment algorithms deals with the average Motion Vector (MV) of the whole available MBs and not pixel based MV recovery. Due to this such algorithms cannot handle the integrity of moving objects/partitions and MVs of available pixels that belong to the estimated object/partition are forced to be identical which may leads to wrong recovery of MVs. Pixel-based Motion Vector with Partition (PMVP) and Spiral Pixel Reconstruction (SPR) algorithms can able to handle multi-directional object movements eventually in temporal domain. Hence, PMVP and SPR algorithms are used as a ground work in this research. The PMVP algorithm predicts MVs of lost macro-blocks (MBs) based on the distance between the lost pixels and the available pixels of the surrounding MBs. The PMVP algorithm is modified by using Mahalanobis distance (MD) rather than Euclidean distance (ED) for better MVs recovery. This modified pixel-based motion vector with partition (MPMVP) gives 1.2% improves in PSNR compared to PMVP. Later, the MPMVP algorithm is upgraded by two different strategies: One by using voting priority of available MVs based on the probabilities of similar directions giving 6.2% improvement in PSNR and second by considering separate horizontal and vertical directions of available MVs in voting priority which gives 8.6% improvement in PSNR at the cost of slight increase in execution time. Similarly, modified spiral pixel reconstruction (MSPR) algorithm based on directional edge recovery method using MD from available pixels of surrounding MBs is developed, leading to 3% and 9% improvement in PSNR compared to existing SPR and PMVP algorithms, respectively.
The motion vectors are generated at the transmitter end in codec-encoder and it is required to reconstruct the frame at the receiving end in codec-decoder. The information available about the MVs at transmitting end can help to find the importance of most repeated MBs in temporal domain. Repeated MBs need to be protected more. Hence there is need of Pre-Transmission Algorithm (PTA) at transmitting end. Now the repeated MBs need not be compressed in PTA. PTA segments video frames into regions of unequal importance with the help of motion vector analysis before coding them independently. PTA with convolutional neural network (CNN) and fuzzy reasoning is developed which will help selecting an appropriate EC for reconstruction. The fuzzy reasoning is supported by Hungarian optimization. The selection of suitable EC technique is done based on quantization parameter, packet loss rate and execution time adaptively. Such an adaptive ECs using PTA gives 20.4%, 3.47% and 6.66% increase in PSNR for “coastguard”, “foreman” and “flower garden” video sequences, respectively with respect to PMVP. The Developed adaptive ECs are also applied on HD standard video data samples. The adaptive EC technique is further generalized so as to be suitable for all types of video codecs, even for HEVC/H.265.
HD/3D video gives a realistic and life like subjective viewing experience and this becomes a major area of research in television broadcasting, video streaming or video transmission technology. To improve existing video standards and its coding efficiency of multi-view video sequences, the Joint Video Team (JVT) introduced multi-view video coding (MVC) that is extended and prolonged by H.264/AVC (Zeng et al., 2014; Yang et al., 2014; Schuster & Katsaggelos 2004; Shuster & Katsaggelos 2006; Zhao et al., 2013; Cheng et al., 2002).