漸進式三維多項式參數視訊編碼之無線傳輸應用

Abstract

為了克服無線通訊的低頻寬限制和增加視訊信號抵抗雜訊的能力，我們在一個名為┌三維多項式參數視訊編碼」(3-D PAC)的視訊壓縮方法上利用空間及時間的切割去達到低位元率的視訊傳輸。藉著結合空間與時間上的切割方式，我們可以在有頻寬限制的環境達到有效且適當的低位元率傳輸3-D PAC的資料。Wiener內插法和線性內插法被用來增加解碼後的影像品質，而且此方法可以維持一定的品質在低位元率的傳輸下。藉著編碼簿的幫助更可以增加壓縮率，使得影像在CIF的格式下傳輸率可低於每秒40千位元，在QCIF的格式下更可低於每秒10千位元。比起相同技術下的H.261傳輸率還低甚至PSNR還比H.261高上1dB。 同時我們模擬了3-D PAC編碼後的影像在無線傳輸的狀況，在不同的調變技術之下，再經過雜訊的通道傳輸後，此3-D PAC編碼方式的PSNR還比H.261高上3dB左右。而且我們也發現合成此編碼方式的硬體實現比起MPEG2的硬體還節省約40%，PSNR高出2~3dB，壓縮率比其高出甚多。

To overcome the bandwidth limitation and to strengthen the channel effect immunity, we use Spatial/Temporal decimation on 3-D PAC for very-low-bit-rate video transmission. After combing the decimation in space and time domain, it is also proved that 3-D PAC is effective and suitable in very low bit rate applications in the environment with limited bandwidth. Wiener interpolator and linear interpolator are applied for quality enhancement. The visual quality can be retained under low bit rate. With the aid of code-book design to gain more compression, 3DPAC would be support the applications with bit rate lower than 40kbits/s in CIF format and 10kbits/s in QCIF format. Compared with H.261 in the same condition, the 3-D PAC is higher 1db than H.261 for PSNR and is lower than H.261 for transmission rate, respectively. We simulate the transmission of 3-D PAC coded video sequence over wireless environment with different modulations. After the transmission in the AWGN channel, it gains advantage on resisting channel noise to 3 dB that compared with videophone standard H.261. And the hardware complexity of 3-D PAC is less 40% than general MPEG2 implementation, while the PSNR of 3-D PAC is even superior to that of MPEG2 by 2~3 dB. ABSTRACT ii ACKNOWLEDGEMENT 謝誌 iii CHPATER 1 INTRODUCTION 1 1.1 Very Low Bit Rates Video Communication for Wireless Applications 1 1.2 Outline of Thesis 3 CHPATER 2 SPATIAL/TEMPORAL DECIMATION ON 3-D PAC CODEC 5 2.1 Algorithm 5 2.1.1 Spatial/Temporal Decimation Model 7 2.1.2 Coefficient Generator 9 2.1.3 Cube Prediction 11 2.1.4 Thresholding 13 2.1.5 Run Length Cube Coding 13 2.1.6 Vector Code-Book Design 14 2.1.7 Interpolation 14 2.2 Architecture 17 2.2.1 Coefficient Generator 17 2.2.2 Qunatization and Inverse Qunatization 19 2.2.3 Thresholding 19 2.2.4 Cube Prediction 20 2.2.5 Run Length Cube Coding 21 2.2.6 Packet Composer 22 CHAPTER 3 SPATIAL/TEMPORAL DECIMATION ON 3-D PAC FOR VERY LOW BIT RATES VIDEO TRANSMISSION 23 3.1 Software Simulation 23 3.2 Simulation for Video Transmission with Channel Noise 26 3.2.1 Simulation Environment 27 3.2.2 COSSAP Simulation Results and Data Analysis 30 CHAPTER 4 HARDWARE IMPLEMENTIONS & PERFORMANCE EVALUATIONS 35 4.1 Architecture Design 35 4.2 Hardware Implementation 36 4.2.1 Coefficient Generator 36 4.2.2 Cube Prediction 38 4.2.3 Thresholding 41 4.2.4 Run Length Cube Coding 42 4.3 Noise Error Analysis 43 CHPATER 5 CONCLUSION & FUTURE WORK 45 REFERENCES 46 APPENDIX A: SEQUENCE OF 3-D PAC 48 APPENDIX B: PUBLICATION 51