Research on Video Coding and Transmission Technology in Wireless Channel
|School||Xi'an University of Electronic Science and Technology|
|Course||Information and Communication Engineering|
|Keywords||H.264 rate control multi-frame motion estimation CABAC unequal protection|
Today, the support of multimedia (including speech, image and video etc.) services in wireless channel or network has become a prosperous research field. However, there exist two challenging problems in wireless video delivery. One is high efficient compression of source for fully utilization of channel bandwidth; the other is protection of compressed information from errors or packet loss. With the improvement of technology, there is an ongoing global trend to shift multimedia applications from traditional platforms, such as desktop PCs and set-top boxes, to hand-hold devices (e.g. PDAs and smart-phones), which results in some new constraints (e.g. power efficient and cost efficient) for VLSI implementation of video codec. In this thesis, we focus on video compression algorithm, VLSI implementation and error control for streaming video in wireless channel.The main contributions and innovation points of the thesis are as follows:1. A PSNR-based basic unit rate control method for H.264 is proposed. Through analysis and study of the rate control in H.264, a peak signal noise ration based (PSNR-based) complexity estimation is proposed, by adding the mean absolute difference based (MAD-based) complexity estimation, to predict the basic unit complexity, by this means more accurate target bits allocation can be achieved. The high computational complexity of the linear model for MAD prediction is a drawback, using the space and temporal correlation among the sequences, we present an algorithm named the weighted model to predict the MADs of the basic unit, which avoids the process of updating the parameters and reduces the computational complexity. The precision of the weighted model is comparable to that of the linear model.2. A novel fast multi-frame motion estimation algorithm based on H.264 is proposed. The proposed algorithm can reduce the computational complexity by developing the algorithm of the initial search point prediction of the reference frames and extending the diamond search to multiple reference frames. The proposed algorithm can achieve an effective integrated trade-off analysis to reach the optimal performance compared with previous methods. Therefore, it is suitable for hardware implementation.3. An efficient hardware implementation of Context-based Adaptive Binary Arithmetic Coding (CABAC) decoder for H.264 is presented. Based on the full use of parallel architecture, an efficient solution for VLSI implementation is described. By developing the two-level finite state machines to control the decoding process and adopting memory clear schedule to solve the problem of coefficients’ time-consuming storage, we can reduce the complexity of CABAC-decoder implementation and increase the speed so as to generate one to two bit within one cycle. Simulation results testify that our design can meet the needs of decoding H.264 main profile 4CIF bit stream at 30fps in real time.4. An adaptive Unequal Loss Protection (ULP) scheme for scalable video streaming over wireless channel is proposed. Based on the available network bandwidth and the packet loss rate estimation, unequal amounts of FEC protection are assigned to each layer and the sending rate is determined by dynamically selecting the number of layers to transmit. To restrain the video transmission error propagation, the FEC assignment algorithm for the base layer is developed by exploring the unequal importance existing in different frames of a group of pictures (GOP), and unequal amounts of protection are allocated to different base layers of a GOP in order to minimize the decoder distortion. The experiment results show that the proposed scheme can largely improve the error resilience of the scalable video bit stream and the performance of video streaming network transmission system.All of the proposed algorithms in this thesis are simulated and examined, then compared with conventional schemes in order to prove their validities and advantages.