Efficient two-pass rate control scheme based on adjusting distribution of discrete cosine transform coefficients

• Published in: IET image processing Vol. 4; no. 3; pp. 211 - 222
• Main Authors: Kuo, C.-H., Chang, L.-C., Chiu, H.-C., Liu, B.-D.
• Format: Journal Article
• Language: English
• Published: Stevenage Institution of Engineering and Technology 01.06.2010; The Institution of Engineering & Technology
• Subjects: Adjustment; Applied sciences; Coding; Coding, codes; Control systems; Detection, estimation, filtering, equalization, prediction; Discrete cosine transform; Distortion; Exact sciences and technology; Glass transition temperature; Image processing; Information, signal and communications theory; Mathematical models; Quadratic programming; Sampling, quantization; Signal and communications theory; Signal processing; Signal, noise; Telecommunications and information theory; Visual; Video coding; Image processing; Data compression; Flow rate regulation; Teletraffic; Image compression; Algorithm; Video signal processing; Variable bit rate; Image quality; Image coding; Group of pictures; Signal quantization; Traffic management; Traffic control; Signal processing; Discrete cosine transforms
• ISSN: 1751-9659; 1751-9667
• DOI: 10.1049/iet-ipr.2009.0083

Video compression systems use rate control (RC) to achieve high visual quality under the constraint of a target bitrate. This paper develops a two-pass variable bit rate (VBR) coding algorithm based on a group-of-picture (GOP) level ρ-domain model. The discrete cosine transform (DCT) coefficients distribution that is generated in the first pass is modified to select an appropriate quantisation parameter (QP) that is used in the second pass, according to the target bit budget of GOP. Besides, the other important parameters, such as the header bits, are also predicted in the second pass to improve the RC ability. If a scene change happens in the first pass coding, the QP value will be adjusted in the second pass coding. Then, the QP is refined based on the textural complexity of every macroblock. The experimental results demonstrate that the proposed VBR methods can not only achieve excellent visual quality but also reduce the variation in distortion even if a scene change happens.