PEG Coding Algorithm Optimization Based On DCT And APBT

JPEG is a popular still image compression standard based on Discrete Cosine Transform (DCT). It has spurred a wide-ranging usage of JPEG format such as in World-Wide-Web and digital cameras. The popularity of the JPEG coding system has motivated the study of JPEG optimization schemes which remain faithful to the JPEG syntax. Since these JPEG optimization schemes have their limitations more or less, there is still a room to further improve the performance of JPEG coding system.Since the conventional DCT-based JPEG compression scheme (DCT-JPEG) is a block-based image compression algorithm, obvious blocking artifacts are remained in reconstructed image at low bit rates, and it adopts different quantization steps for different DCT coefficients which makes the quantization table more complex. All Phase Biorthogonal Transform (APBT)-based JPEG compression scheme (APBT-JPEG) is a good solution to these problems. However, to the best of our knowledge, there is still no fast algorithm for computing APBT which causes higher computational complexity in the transform step of APBT-JPEG. Thus, APBT-JPEG cannot meet the requirement of real-time image processing. After a comparative study of DCT-JPEG and APBT-JPEG, we propose two JPEG optimized schemes to further improve the performance of JPEG scheme. These JPEG optimized schemes are Windowed All Phase Biorthogonal Transform-based JPEG (WAPBT-JPEG) and a novel optimized quantization table for DCT-JPEG. The principal innovations of this thesis are as follows:(1) In the light of Windowed All Phase Digital Filter (WAPDF) theory, we propose new concepts of the Windowed All Phase Biorthogonal Transform (WAPBT), which is inspired by the APBT. The matrices of WAPBT based on DFT, WHT, DCT, IDCT, and DST are deduced and the properties of WAPBT are studied. The WAPBT is considered to be APBT when rectangular window is choosed as its window sequence. Therefore, with reference to the high performance of APBT-JPEG, we propose a JPEG optimization scheme called WAPBT-based JPEG (WAPBT-JPEG) in order to further improve the performance of JPEG coding system. With the optimal window sequence of WAPBT for JPEG compression codec obtained by using the optimization technique, the application of WAPBT in JPEG coding system is realized.(2) With our in-depth study on APBT and APBT-JPEG, we have found the relation between APBT-JPEG and DCT-JPEG. The relation is that APBT-JPEG can be implemented by DCT-JPEG by using a new quantization table deduced from mathematical analysis and the corresponding experiment results have verified this relation both in terms of objective quality and subjective effect. Based on this, we propose a novel quantization table used in DCT-JPEG in order to avoid extra computational complexity in APBT-JPEG and an almost identical coding performance is achieved compared with APBT-JPEG. Tested by natural images, experimental results show that at low bit rates, by using our proposed quantization table in DCT-JPEG, blocking artifacts in reconstructed image have been reduced significantly.Since the proposed schemes in this thesis only optimize the JPEG codec without introducing substantial modifications, they can not only further reduce the sizes of JPEG compressed images, but also have the advantage of being easily deployable, such as in the application of wireless communications.