TY - JOUR
T1 - A new low-complexity approximate DCT for image and video compression
AU - Lin, Hsin Kun
AU - Sun, Chi Chia
AU - Sheu, Ming Hwa
AU - Berecovic, Mladen
N1 - Funding Information:
The authors are very grateful to the anonymous referees for their detailed comments and suggestions regarding this paper. This research was supported in part by the Ministry of Science and Technology of Taiwan, under the contracts of MOST-107-2911-I-150 -501, Department of Electrical Engineering, National Formosa University, Smart Machine and Intelligent Manufacturing Research Center, National Formosa University, Center for Doctoral Studies Lübeck, Universität Zu Lübeck.
Publisher Copyright:
© 2020 The Chinese Institute of Engineers.
Copyright:
Copyright 2020 Elsevier B.V., All rights reserved.
PY - 2020/8/17
Y1 - 2020/8/17
N2 - In this paper, a new approximate discrete cosine transform (DCT) architecture for an H.265/HEVC video compression codec is proposed. In order to reduce the computational complexity, a configuration of basic elements by 0, ±0.5, ±1 is proposed to improve the DCT coefficient performance. Next, we derive different DCT coefficients for different transform block sizes based on DCT symmetry properties for H.265. Experimental results show that the proposed DCT can increase by 4 dB with a peak signal-to-noise ratio (PSNR) on average compared with general approximate DCT approaches. In comparison with recent approximation DCTs, the proposed architecture cannot only reduce the computational complexity but can also obtain good quality in PSNR. The proposed configurable DCT architecture successfully reduces approximately 40% of the hardware cost for multiple transform blocks.
AB - In this paper, a new approximate discrete cosine transform (DCT) architecture for an H.265/HEVC video compression codec is proposed. In order to reduce the computational complexity, a configuration of basic elements by 0, ±0.5, ±1 is proposed to improve the DCT coefficient performance. Next, we derive different DCT coefficients for different transform block sizes based on DCT symmetry properties for H.265. Experimental results show that the proposed DCT can increase by 4 dB with a peak signal-to-noise ratio (PSNR) on average compared with general approximate DCT approaches. In comparison with recent approximation DCTs, the proposed architecture cannot only reduce the computational complexity but can also obtain good quality in PSNR. The proposed configurable DCT architecture successfully reduces approximately 40% of the hardware cost for multiple transform blocks.
UR - http://www.scopus.com/inward/record.url?scp=85088819700&partnerID=8YFLogxK
U2 - 10.1080/02533839.2020.1786463
DO - 10.1080/02533839.2020.1786463
M3 - Journal articles
AN - SCOPUS:85088819700
SN - 0253-3839
VL - 43
SP - 580
EP - 591
JO - Journal of the Chinese Institute of Engineers, Transactions of the Chinese Institute of Engineers,Series A
JF - Journal of the Chinese Institute of Engineers, Transactions of the Chinese Institute of Engineers,Series A
IS - 6
ER -