UNCORRECTED PROOF Adaptive Management of Video Encoder Complexity I n this paper, we investigate methods of reducing the computational complexity of the discrete cosine transform (DCT) in a software video encoder. The number of DCT calculations may be reduced by modeling the distribution of zero blocks. We demonstrate that the reduction in computational complexity is variable and depends on the statistics of the video sequence. We propose a new adaptive algorithm that can maintain a near-constant reduction in complexity. The proposed algorithm performs well at converging to a ‘‘target’’ computational complexity, at the expense of a small reduction in image quality. This algorithm provides a flexible mechanism for managing computational complexity in a video encoder. # 2002 Elsevier Science Ltd. Iain E.G. Richardson and Yafan Zhao School of Electronic and Electrical Engineering, The Robert Gordon University, Aberdeen AB10 1FR, UK E-mail i.g.richardson@rgu.ac.uk Introduction There is a growing market for video communication applications that incorporate software-only implemen- tations of popular video coding standards such as H.263 [1] and MPEG4 [2]. These include conferencing and streaming video applications on the PC platform as well as emerging applications for embedded processors and digital signal processors [3]. In many cases, the performance of these software- only CODECs is limited by available processing power as well as, or rather than, by available bandwidth. This is particularly true for applications using low-power embedded processors or in cases where the video CODEC must compete for processor resources with several other applications. It is therefore important to develop flexible methods of managing the computa- tional complexity of video encoding and decoding. In this paper, we take the following approach to support complexity management for real-time video encoding. The discrete cosine transform (DCT) [4] is identified as a computationally intensive function. Methods of predicting or modeling the output of the DCT (and therefore bypassing some computational steps) are compared. An adaptive algorithm is described and is shown to be suitable for dynamically controlling the complexity of the DCT (and related functions) to maintain a ‘‘target’’ level of computational complexity. DCT Complexity Popular video coding standards such as H.263 and MPEG4 make use of DCT encoding of motion- compensated residual frames. Within this type of video CODEC, encoding requires more processing power than decoding (due to the extra computation required for 1077-2014/02/000000+00 $35.00/0 r 2002 Elsevier Science Ltd. Real-Time Imaging 8, 000–000 (2001) doi:10.1006/rtim.2001.0275, available online at http://www.idealibrary.com on YRTI: 275 7.51c/W (Jun 11 2001) SG Dorthy. A Venki