| Summary: | 碩士 === 義守大學 === 電子工程學系碩士班 === 98 === H.264 is the newest video coding standard and it can achieve higher compression efficiency compared to previous standards (such as MPEG-4 and H263) in the same video quality. To serve diversified clients over heterogeneous network, the new scalable video coding (SVC) standard is developed by the Joint Video team (JVT) of the ITU-T and the ISO/IEC as an amendment of H.264. The SVC can provide a full scalability including spatial, temporal and signal-to-noise ratio (SNR) scalability. But in the wireless communication network''s environment, the video quality of H.264 SVC destroyed by channel noise is very severe. In order to improve the performance when H.264 SVC is applied in the wireless channel, the thesis develops a two-dimensional unequal error protection (2-D UEP) for the scalable video with a combined temporal and quality (SNR) scalability over packet-erasure channel. We derive the bit-allocations by combining the source and channel encoder for base layer and enhancement layers, respectively.
In order to avoid the waste of bits and achieve the best bit-allocation, our thesis first discusses two methods for the PSNR bit-allocation algorithm (PSNR-BAA) [9-10] and layer weight bit-allocation algorithm (LW-BAA) [11]. PSNR-BAA encodes the video sequence by using unequal protection technique. The bit-allocation is obtained by the PSNR-BAA according to the PSNR increment of the related layers. Since it need to compute PSNR for every layers, the computation load of the PSNR-BAA is very huge. To improve the defect of PSNR-BAA, the LW-BAA uses a simple layer weight to allocate the bit-allocation matrix. The simple mathematic model derived by [10] can achieve the effect of acceleration, but it doesn’t consider the effect of packet length and error correct of RS code. It leads to allocate too much protect code used in the base layer. To improve the defects of above two methods, we propose a threshold-based bit-allocation algorithm (TH-BAA). The proposed TH-BAA designs a predefined threshold according to the length of packet and the ability of RS code error correcting to avoids the waste of bits. In addition, the proposed TH-BAA also derives a simple mathematical model to reduce computation load of bit allocation.
Form the experimental results, we can find the decoded images of the PSNR-BAA is better when compared to PSNR-BAA and LW-BA using Foreman sequence under the bandwidth is 100 KB/s and packet loss rate is 30%. The TH-BAA alleviates the drops of PSNR. We can get the same performance for Bus sequence in the 140 KB/s of bandwidth. The PSNR improvement of TH-BAA is better than PSNR-BAA and LW-BAA about 2~6 dB of average. In addition, the PSNR improvement for the Akiyo sequence is approximately 8 dB under 50 KB/s.
|
|---|
