本論文提出了用於低位元率的控制編碼演算法，利用此演算法可調整輸出位元率並且維持影像的品質。這個演算法是適合於典型視訊會議、視訊電話等低位元率的影像編碼應用。首先，將要編碼的影像序列作分析,找出每張影像每個區塊的主要屬性。接著我們利用這些屬性來決定量化參數。換句話說，在決定量化參數時，不只要考慮緩衝區的狀態還要考慮編碼影像的特性。最後我們將在視訊標準H.263下實現我們的方法。模擬結果顯示我們提出的演算法優於傳統的方法。

In this thesis, coding control algorithms for very low bit rate video are proposed to regulate the output data rate and maintain the output video quality. It is suited for very low bit rate coding of "head-and-shoulders" video material typical of video teleconferencing, and video telephony. In this research, the content of the coded picture is first analyzed to find the main attributes of each macroblock. The attributes of the macroblocks are then taken into account while adjusting the quantization parameters. In other words, the quantization parameters are adjusted not only depending on the fullness of the output buffer but also on the content of the coded image. The methods are implemented on H.263 which is a video coding standard for narrow band telecommunication channels up to 64kbit/s recommended by ITU-T SG 15, to show the effectiveness. The simulation result show that the perceptual quality of the coded pictures, by using the proposed methods, is better while comparing with the traditional approach.

TABLE OF CONTENTS
Page
TABLE OF CONTENTSi
LIST OF FIGURESiii
ABSTRCAT vi
CHAPTER 1. INTRODUCTION1
1.1 Background1
1.2 H.263 Video Coding System3
1.3 Outline of This Thesis8
CHAPTER 2. H.263 CODING CONTROL REVIEW11
2.1 Introduction11
2.2 H.263 TEST MODEL (TMN5) Rate-Control Mechanism17
CHAPTER 3. PREPROCESSING FOR CODING CONTROL18
3.1 Introduction18
3.2 Overview of the Proposed System Structure20
3.2.1 Extraction of Active Regions21
3.2.2 Prefiltering25
3.3 Summary27
CHAPTER 4. RATE CONTROL ALGORITHMS FOR VERY LOW BIT RATE CODING28
4.1 Rate Control Scheme28
4.1.1 MAD-based rate control methodⅠ30
4.1.2 MAD-based method Ⅱ32
4.1.3 Activity-based method34
4.2 Simulation Results35
CHAPTER 5. CONCLUSIONS.54
REFERENCES56
LIST OF FIGURES
Page
Figure 1.1 H.263 Source Coder9
Figure 1.2 The Macroblock defined for H.263 coding system9
Figure 1.3 H.263 Syntax diagram for the video bitstream10
Figure 2.1 Block diagram of simple backward buffer control scheme13
Figure 3.1 The block diagram of the traditional video coding system19
Figure 3.2 The block diagram of the proposed video coding system19
Figure 3.3 Sobel operators: (a) 3×3 image mask region; (b)mask used to compute Gx at center point of the 3×3 region; (c) mask used to compute Gy at that point.22
Figure 3.4 The original current image24
Figure 3.5 The original previous image24
Figure 3.6 The gradient image obtained from the current image25
Figure 3.7 The gradient image obtained from the previous image25
Figure 3.8 The detected active pixels in the current image25
Figure 3.9 The active marcoblocks in the current image25
Figure 3.10 A 3×3 FIR prefilter26
Figure 3.11 Content-based Prefiltered Image27
Figure 4.1 Coding rate control procedure by using the MAD-based methodⅠ33
Figure 4.2 The PSNR comparison at 14.4kbps for image sequence "akiyo"37
Figure 4.3 The PSNR comparison at 28.8kbps for image sequence "akiyo"37
Figure 4.4 The PSNR comparison at 33.6kbps for image sequence "akiyo"38
Figure 4.5 The PSNR comparison at 14.4kbps for image sequence "irene"38
Figure 4.6 The PSNR comparison at 28.8kbps for image sequence "irene"39
Figure 4.7 The PSNR comparison at 33.6kbps for image sequence "irene"39
Figure 4.8 The PSNR comparison at 14.4kbps for image sequence "Miss America"40
Figure 4.9 The PSNR comparison at 28.8kbps for image sequence "Miss America"40
Figure 4.10 The PSNR comparison at 33.6kbps for image sequence "Miss America"41
Figure 4.11 The original image of "akiyo"42
Figure 4.12 The reconstructed image of "akiyo" at 14.4kbps with TMN542
Figure 4.13 The reconstructed image of "akiyo" at 28.8kbps with TMN5r42
Figure 4.14 The reconstructed image of "akiyo" at 33.6kbps with TMN542
Figure 4.15 The original image of "akiyo"43
Figure 4.16 The reconstructed image of "akiyo" at 14.4kbps with MAD-based methodⅠ43
Figure 4.17 The reconstructed image of "akiyo" at 28.8kbps with MAD-based methodⅠ43
Figure 4.18 The reconstructed image of "akiyo" at 33.6kbps with MAD-based methodⅠ43
Figure 4.19 The original image of "akiyo"44
Figure 4.20 The reconstructed image of "akiyo" at 14.4kbps with MAD-based methodⅡ44
Figure 4.21 The reconstructed image of "akiyo" at 28.8kbps with MAD-based methodⅡ44
Figure 4.22 The reconstructed image of "akiyo" at 33.6kbps with MAD-based methodⅡ44
Figure 4.23 The original image of "akiyo"45
Figure 4.24 The reconstructed image of "akiyo" at 14.4kbps with activity-based method45
Figure 4.25 The reconstructed image of "akiyo" at 28.8kbps with activity-based method45
Figure 4.26 The reconstructed image of "akiyo" at 33.6kbps with activity-based method45
Figure 4.27 The original image of "irene"46
Figure 4.28 The reconstructed image of "irene" at 14.4kbps with TMN546
Figure 4.29 The reconstructed image of "irene" at 28.8kbps with TMN546
Figure 4.30 The reconstructed image of "irene" at 33.6kbps with TMN546
Figure 4.31 The original image of "irene"47
Figure 4.32 The reconstructed image of "irene" at 14.4kbps with MAD-based methodⅠ47
Figure 4.33 The reconstructed image of "irene" at 28.8kbps with MAD-based methodⅠ47
Figure 4.34 The reconstructed image of "irene" at 33.6kbps with MAD-based methodⅠ47
Figure 4.35 The original image of "irene"48
Figure 4.36 The reconstructed image of "irene" at 14.4kbps with MAD-based methodⅡ48
Figure 4.37 The reconstructed image of "irene" at 28.8kbps with MAD-based methodⅡ48
Figure 4.38 The reconstructed image of "irene" at 33.6kbps with MAD-based methodⅡ48
Figure 4.39 The original image of "irene"49
Figure 4.40 The reconstructed image of "irene" at 14.4kbps with activity-based method49
Figure 4.41 The reconstructed image of "irene" at 28.8kbps with activity-based method49
Figure 4.42 The reconstructed image of "irene" at 33.6kbps with activity-based method49
Figure 4.43 The original image of "Miss American"50
Figure 4.44 The reconstructed image of "Miss American" at 14.4kbps with TMN550
Figure 4.45 The reconstructed image of "Miss American" at 28.8kbps with TMN550
Figure 4.46The reconstructed image of "Miss American" at 33.6kbps with TMN550
Figure 4.47 The original image of "Miss American"51
Figure 4.48 The reconstructed image of "Miss American" at 14.4kbps with MAD-based methodⅠ 51
Figure 4.49 The reconstructed image of "Miss American" at 28.8kbps with MAD-based methodⅠ 51
Figure 4.50 The reconstructed image of "Miss American" at 33.6kbps with MAD-based methodⅠ 51
Figure 4.51 The original image of "Miss American"52
Figure 4.52 The reconstructed image of "Miss American" at 14.4kbps with MAD-based methodⅡ 52
Figure 4.53 The reconstructed image of "Miss American" at 28.8kbps with MAD-based methodⅡ 52
Figure 4.54 The reconstructed image of "Miss American" at 33.6kbps with MAD-based methodⅡ 52
Figure 4.55 The original image of "Miss American"53
Figure 4.56 The reconstructed image of "Miss American" at 14.4kbps with activity-based method 53
Figure 4.57 The reconstructed image of "Miss American" at 28.8kbps with activity-based method 53
Figure 4.58 The reconstructed image of "Miss American" at 33.6kbps with activity-based method 53

[1] ITU SGXV Recommendation H.261, "Video Coder for Audiovisual Services at p*64 kbits ".
[2] ISO/IEC MPEG, "Coding of moving pictures and associated audio". Committee Draft of Standard ISO11172, Dec. 1991.
[3] ITU-T Recommendation H.262, ISO/IEC 13818-2, "Generic coding of moving pictures and associated audio information". Draft International Standard, Mar. 1995.
[4] Draft ITU-T Recommendation H.263, "Video coding for low bit rate communication". July 1995.
[5] ISO/IEC JTC/SC29/WG11 N1727,"MPEG-4 requirements document V.4," July 1997
[6] K. H. Tzou, "An intrafield DCT-Based HDTV coding for ATM networks," IEEE Trans. On Circ. And Syst. For Video Technol., vol. 1, pp. 184-196, Jun. 1991.
[7] A. Puri and R. Aravind, "Motion-compensated video coding with adaptive perceptual quantization,: IEEE Trans. On Circ. And Syst. For Video Technol., vol. 1, pp. 351-361, Dec. 1991.
[8] E. ViScito and C. Gonzales, "A video compression algorithm with adaptive bit allocation and quantization, " SPIE Proc: Visual Communications and Image Processing, vol. 1605, pp. 58-72, Nov. 1991.
[9] ISO/MPEG Ⅱ, "Test Model 5," Document AVC-491, Apr. 1993.
[10] J. Zdepski, D. Raychaudhuri and K. Joseph, "Statistically based buffer control policies for a constant rate transmission of the compressed digital video," IEEE Trans. On Commu., vol. 39, pp.947-957, Jun. 1991
[11] CCITT SG XV WP/1/Q4 Specialists Group on Coding for Visual Telephony, "Description of Reference Model 8 (RM8), June 1989, Document 525.
[12] ITU Telecommunication Standardization Sector LBC-95, Study Group 15, Workin Party 15/1. Expert's Group on Very Low Bit rate Visual Telephony. VIDEO CODEC TEST MODEL, TMN5.
[13] W. Tawbi, f. Horn, E. Horlait, and J-B . Stefani, "Video Compression Standards and Quality of service," The Computer Journal, vol 36, no. 1, pp. 43-54, Jan 1993.
[14] Y. Kim and H. Kim, "Anew Buffer Control Strategy for Image Data Compression," IEEE Trans. On Consumer Electronics, vol. 40, pp. 932-937, Nov. 1994.
[15] M. T. Sun, K. Rijkse, D Schinkel and A meijboom, "Coding and Interworking for Videotelphony," Proc. ISCAS-93, pp.20-23, 1993.
[16] Karl Olav Lillevold, "The software of a very low bit rate video encoder producing H.263 bitstreams.", Telenor Research and Development, June 14, 1996.
[17] Rafael C. Gonzalez, and Richard E. Woods, "Digital image processing," Addision-Wesley Publishing Company.
[18] H. Y. Gong and H. M. Hang, "Scene analysis for DCT image coding," Signal Processing of HDTV, V, pp. 425-434, 1994