高吞吐率JPEG2000编码器VLSI设计  

Design of a hardware compression encoder with a high throughput

作　　者：吴长坤 张为 郝亚喆 WU Changkun;ZHANG Wei;HAO Yazhe(School of Microelectronics,Tianjin University,Tianjin 300072,China)

机构地区：[1]天津大学微电子学院,天津300072

出　　处：《西安电子科技大学学报》2022年第4期176-183,共8页Journal of Xidian University

基　　金：光电信息控制和安全技术重点实验室项目(JCKY2019210C053);天津市科技计划(20YDTPJC00890)。

摘　　要：针对实际应用中海量图像数据造成的存储与传输带宽的压力,设计了一种兼容处理8~16位灰度图的高吞吐率图像压缩编码器,并给出了相应的超大规模集成电路架构。通过分析16位灰度图像像素值的特点,以及JPEG2000压缩编码器在硬件实现中,离散小波变换与优化截断嵌入式模块处理时间的差异,提出了并-串-并的优化结构,架构采用高8位和低8位处理进行设计,高位和低位结构既可以独立进行处理,又可以结合起来共同处理,增加了编码器的灵活性,大大提高了压缩编码器的吞吐率。编码器最终在Xilinx XC7K480T芯片上实现,最高工作频率可达147.734 MHz,对8位典型灰度图进行处理时,得到的最大吞吐率为169.55 MB/s;对16位典型灰度图进行处理时,最大吞吐率高达266.87 MB/s。与现有同类编码器相比,吞吐率提高40%以上。在实际工程应用中,该编码器不仅可靠性高、灵活性好,而且可扩展性强,通过控制并行度可实现不同分辨率图像的高倍率、高质量快速压缩,具有重要的应用价值。Aiming at the pressure of storage and transmission bandwidth massive image data in practical applications,a high throughput image compression encoder compatible with processing 8~16 bits grayscale images is designed,and the corresponding VLSI architecture is given.By analyzing the characteristics of the 16-bit gray image pixel value,and the difference between Discrete Wavelet Transform(DWT)and Optimized Truncated Embedded Module(EBCOT)with the hardware implementation of the JPEG2000 compression encoder in the processing time,the optimal parallel-series-parallel structure is proposed,and the architecture is designed with high 8-bit and low 8-bit processing,which can be processed independently or together.This architecture increases the flexibility of the encoder and greatly improves the throughput of the compression encoder.Finally,the encoder is implemented on Xilinx XC7 K480 T.The highest operating frequency of the encoder is 147.734 MHz,and the maximum throughput rate of the 8-bit grayscale image is 169.55 MB/s.The maximum throughput of the 16-bit grayscale image is 266.87 MB/s.Compared with existing similar encoders,the throughput is increased by more than 40%.In practical engineering applications,the encoder has not only a high reliability and good flexibility,but also a strong scalability.By controlling the parallelism,it can realize high magnification,high quality and fast compression of images with different resolutions,which is of important practical value.

关 键 词：图像压缩 超大规模集成电路 并行处理 吞吐量 

分 类 号：TN47[电子电信—微电子学与固体电子学]