Merged inverse quantization and IDCT for optimized decoder implementation

TL;DR: The proposed method merges the inverse quantization and IDCT into a single procedure, which is referred to as inverse quantized IDCT (IQIDCT), such that the decoded coefficients do not need to be inverse quantizing prior to the inverse transform.

Abstract: This paper proposes an efficient technique to reduce the complexity of implementation for inverse discrete cosine transform (IDCT)-based video decoders. The proposed method merges the inverse quantization and IDCT into a single procedure, which is referred to as inverse quantized IDCT (IQIDCT), such that the decoded coefficients do not need to be inverse quantized prior to the inverse transform. Thus, the computations related to inverse quantization are omitted. Since the cosine transform basis can be constructed before decoding, the proposed IQIDCT effectively replaces the need for computing power with little additional memory. The performance of the proposed algorithm is evaluated by comparing it with the original XVID decoder which uses the traditional separate inverse quantization and IDCT method. Experiments show the efficiency of the proposed method in reducing the overall decoding complexity. Moreover, it does not result in any video quality degradation. The proposed method is particularly suitable for low-power processors in multimedia systems.