Study On Key Techniques In Wireless Video Dilivery

The emerging new techniques in multimedia and communication fields lead to thetremendous change in the way and content of personal communication. In particular, with thefoundation of 3G and beyond, wireless video delivery has been the focus both in wirelesscommunication and multimedia field. In this paper, we deal with several critical techniques inwireless video delivery (mainly refers to mobile communication) and make severalcontributions:1. Rate control based on statistical features of original video. Since the bandwidth isrelatively low in wireless transmission whereas video sequence has large amount of data,rate control scheme optimized for low bit rate video delivery is essential. In the paper, wehave analyzed the existed rate control schemes and found that there was a commonassumption in them, i.e., the bit rate variation of a video sequence could be modeled by asteady random process (in which the statistical parameters can be estimated by previousencoded data). Such an assumption is not accurate in low bit rate video coding since themotion data makes up a large part of bitstream, which cannot be estimated accuratelyenough by a single steady random process due to the fact that the object motion in videosequence is unsteady. Thus the existed methods cannot track the variation of bit rate in lowbit rate video coding in time. In order to solve the problem, a new bit allocation scheme isproposed based on statistical motion feature of original video, in which bits assignment isparallel to the degree of motion in a picture. At the basis of this, a quantization parameter(QP) estimation scheme is presented by a simple encoding complexity prediction method.Besides, to improve the R-D performance of encoding further, a R-D optimal I-frame QPestimation scheme is given. The final simulation results show that the proposed scheme iseffective.2. Error resilient transmission scheme based on minimal bandwidth-distortion (B-D)cost. The concept of B-D cost is proposed to evaluate a wireless video delivery system.With the concept, a partial retransmission scheme is presented, in which we develop anadaptive error recovery mode selection algorithm (here the error recovery mode refers toARQ and error concealment by post-processing). As is known, a main defect in wirelesstransmission is low bandwidth. When compressed video data are corrupted at receiver side,there are two approaches which can be applied to recover the erroneous data: request forretransmission by ARQ or error concealment by spatial-temporal interpolation. The existedschemes usually try their best to use ARQ and the error concealment is not taken until ARQhas been failed. As a matter of fact, error concealment may obtain recovered video withgood enough quality since there is strong spatial and temporal correlation in videosequence. In this case, using ARQ leads to obvious waste of bandwidth, which is notfeasible to wireless video delivery. To deal with the problem, we use B-D cost to determinewhether the ARQ is used or not for corrupted video data. The final simulation results showthat the proposed scheme is effective.3. Joint source channel coding (JSCC) with arithmetic codes (AC). Error resilienceencoding is one of the key techniques in wireless video delivery. Since the traditionalmethods, in which channel coding and source coding are separately designed, cannotobtain optimal transmission performance in wireless communication, JSCC has been thefocus in recent years. JSCC with AC is a popular branch in JSCC field. However, theexisted methods of JSCC with AC are either of high computational complexity or low errorcorrection capability. Hence, to deal with the high computational complexity in MAPdecoding, we proposed local MAP (LMAP) decoding, in which the computationalcomplexity and error correction capability can be effectively balanced. Based on LMAPscheme, maximal distantly weighted a posteriori probability decoding (MDWAP) ispresented to provide better error correction capability than MAP while keepingcompromised computational complexity.Based on the analysis of the existed approaches, we improve the performance ofend-to-end video delivery by introducing new models or metrics in the paper. In the proposedrate control scheme, new parameter models for bits allocation, QP estimation and I-frame QPare presented based on statistical feature of original video. In the proposed error resilienttechniques (include error resilient transmission and coding), B-D cost metric, LMAP metricand DWAP metric are provided. With these new models and metrics, the proposed approachescan provide better end-to-end video transmission quality, reduce the computational complexityat receiver side and relax the inherent contradiction in wireless video delivery.