集成电路与嵌入式系统2026，Vol.26Issue(6)：35-42,8.DOI:10.20193/j.ices2097-4191.2025.0155
一种基于频域变换优化的视频压缩算法
A video compression algorithm based on frequency domain transform optimization
摘要
Abstract
Aiming at the performance degradation of discrete cosine transform(DCT)in video coding when processing non-stationary re-sidual signals,this paper proposes a frequency-domain transform optimization method based on reversible structural perturbation.With-out modifying the kernel of the standard DCT transform,this method preprocesses the prediction residuals via a set of reversible row cy-clic shift operations to adjust their spatial statistical distribution,making it closer to the optimal working region of DCT and thus impro-ving transform coding efficiency.The algorithm dynamically calculates the offset according to the width and height of the coding unit(CU),and adaptively selects the optimal perturbation mode(including the non-perturbation mode)using a rate-distortion optimization mechanism.The mode indicator is signaled in the bitstream with extremely low overhead,and lossless reconstruction of the residuals is realized at the decoder through the corresponding inverse perturbation.Experimental results based on three full-HD test sequences in Class B of the JCT-VC standard show that the proposed method can achieve an average bitrate reduction of approximately 1.91%in overall tests with mixed CU sizes and prediction modes.The performance gain is especially significant in inter-prediction modes and large-size CUs.Specifically,a bitrate reduction of 2.80%is obtained for 4096-pixel CUs in the ParkScene sequence,and the algorithmic gain exhibits a U-shaped characteristic of"high at both ends and low in the middle"as the CU area increases.The proposed method effectively improves compression performance while maintaining computational complexity similar to that of the standard DCT,provi-ding a new idea for hardware-friendly optimization of the transform module in video coding.关键词
视频编码/离散余弦变换/Karhunen-Loève变换/残差预处理/率失真优化/结构扰动Key words
video coding/discrete cosine transform/Karhunen-Loève transform/residual preprocessing/rate-distortion optimization/structural perturbation分类
信息技术与安全科学引用本文复制引用
姜伟,张博,李国新,张悦,张韵东..一种基于频域变换优化的视频压缩算法[J].集成电路与嵌入式系统,2026,26(6):35-42,8.基金项目
北京学者计划(2023NO.095) （2023NO.095）
