A Novel Deblocking Quantization Table for Luminance Component in Baseline JPEG

摘要

All Phase Biorthogonal Transform (APBT), a DCT-like transform generated from All Phase Digital Filter (APDF), can be used in baseline JPEG by replacing conventional Discrete Cosine Transform (DCT), and the image compression scheme is called APBT-based JPEG (APBT-JPEG). APBT-JPEG can achieve better coding performance but at the expense of an increased computational complexity, because there is no fast algorithm for computing APBT. With our in-depth study on APBT-JPEG, we have found the relation between APBT-JPEG and DCT-JPEG (baseline JPEG) in this paper, hi order to avoid extra computational complexity in APBT-JPEG, we propose a novel quantization table used in DCT-JPEG and an almost identical coding performance is achieved compared with APBT-JPEG. Tested by natural images, experimental results show that compared with other quantization tables, at low bit rates in DCT-JPEG, the performance both in terms of PSNR and visual quality can be improved by using our proposed quantization table, and the blocking artifacts in reconstructed image have been reduced significantly. For these reasons, we can foresee that the proposed quantization table will be widely used in the future.