JPEG2000中DWT-EBCOT联合的高效低存储VLSI结构
doi: 10.3724/SP.J.1146.2007.01637
An Efficient DWT-EBCOT Combined VLSI Architecture with Low Memory for JPEG2000
-
摘要: 针对JPEG2000硬件实现中小波变换与编码之间占用大量存储的问题,该文提出一种基于码块的存储方案。通过对码块大小片内存储最大程度的复用以及对其高效简单的调度控制,从面积和功耗两方面减小了硬件实现的开销。在实现中,采用基于行的提升变换结构和比特平面并行的编码方式,提高了效率,确保整个过程的实时处理。实验结果表明:在实时编码要求下,对分辨率为512512的图像分片进行四级9/7或者5/3小波分解,码块大小为3232,采用本文结构所用的存储量与直接使用外部存储器的方法相比可减少80%以上。整个结构已通过FPGA验证,且系统时钟可以工作在100MHz。Abstract: Motivated by an enormous amount of storage between DWT and EBCOT in JPEG2000 hardware implementation, a novel memory-efficient scheme based on code block size is proposed. Further reuse of on-chip code block size memories and efficient scheduling of them reduce hardware cost in both area and power. In the implementation, line-based lifting DWT architecture and bit plane parallel EBCOT design are used. The resulting coding efficiency is improved and the whole architecture can achieve real time processing. Experimental results show that on demand of real time coding, when a tile with resolution up to 512-width and 512-height is decomposed with four levels in 9/7 or 5/3 filters and the size of code block is 3232, the wavelet coefficients memory needed in the proposed architecture is reduced by over 80%, compared with existing architectures that wavelet coefficients are directly deposited in off-chip memories. The whole design has been synthesized and mapped into Xilinx FPGA, passing the verification when the system is working at the clock of 100MHz.
-
ISO/IEC JTC 1/SC 29/WG N1646R, ISO/IEC FCD15444-1.JPEG2000 Part Final Committee Draft Version 1.0 [S]. 16March 2000.[2]Daubechies I and Sweldens W. Factoring wavelet transforminto lifting steps [J].J. Fourier Anal. Applicat.1998, 4(2):247-269[3]Chiang Jen-Shiun, Lin Yu-Sen, and Hsieh Chang-Yo.Efficient pass-parallel architecture for EBCOT in JPEG2000[C]. Proceedings of IEEE International Symposium Circuitsand Systems, Scottsdale, Arizona, USA, 2002, I: 773-776.[4]Lian Chung-Jr, Chen Kuan-Fu, and Chen Hong-Hui, et al..Analysis and architecture design of block-coding engine forEBCOT in JPEG2000 [J].IEEE Trans. on Circuits andSystem for Video Technology.2003, 13(3):219-230[5]Acharya T and Chakrabarti C. A survey on lifting-baseddiscrete wavelet transform architectures [J]. Journal of VLSISignal Processing, 2006, 42(3): 321-339.[6]Cheung Ngai-Man, Tang C, and Ortega A, et al.. Efficientwavelet-based predictive Slepian-Wolf coding forhyperspectral imagery [J]. Eurasip Journal on Applied SignalProcessing, 2005, 86(11): 3180-3195.[7]Andra K, Chakrabarti C, and Acharya T. A highperformanceJPEG2000 architecture [J].IEEE Trans. onCircuits and System for Video Technology.2003, 13(3):209-218[8]Huang Quanping, Zhou Rongzheng, and Hong Zhiliang. Lowmemory and low complexity VLSI implementation ofJPEG2000 codec [J].IEEE Trans. on Consumer Electronics.2004, 50(2):638-646
计量
- 文章访问数: 3470
- HTML全文浏览量: 89
- PDF下载量: 798
- 被引次数: 0