多核处理器的JPEG2000并行解压缩算法优化设计

发布时间：2018-08-25 12:28

【摘要】：随着对地观测卫星等研制需求迅速增长,需要处理的遥感图像数据量也急剧增大,这对地面图像检测与处理系统的研制提出了严峻的挑战。针对现阶段数据传输通道码率高、数据处理复杂、运算速度快等特点,大数据量遥感图像的高速解码日益受到人们重视。卫星上装载的芯片采用JPEG2000压缩算法,目前该芯片已经能够对拍摄到的图像进行实时压缩,但是地面接收端还未能够实现实时解压缩。本文主要围绕解决这一实际问题进行研究。为了解决JPEG2000压缩标准算法复杂性带来的解码耗时较长的问题,本文研制采用多核处理器来构建高性能JPEG2000实时解码处理系统。多核处理器是把多个完整的计算引擎集成到一颗处理器上,其中每个引擎都可以独立并行计算,相比于传统的单核处理器平台具有很高的性能提升。本文所用的多核处理器平台,是由Tilera公司生产的4颗TILE-GX36处理核心组成,每颗处理核心拥有36个独立内核。为了实现快速可靠的解压缩图像,本文将JPEG2000解码算法嵌入该多核处理器平台,并且针对解码算法以及平台的调度机制进行了有效的优化,最终解决了当前任务中解压缩实时速率不高的紧迫问题。本文的正文部分首先概述了JPEG2000算法以及多核处理器的发展历史及现状,并简单介绍了多核处理器的核心技术;然后详细的论述了JPEG2000编解码核心算法;最后,本文提出并实现了基于多核处理器Tilera平台的JPEG2000解码高速并行方案,该并行方案不仅体现在多个处理器内核之间,而且体现在一颗处理器内核上的36个更小的执行核心之间。此外,本文还针对解码算法以及多核平台调度方面进一步优化,从而大大提升了解码速度。最后基于Tilera平台进行了JPEG2000解码性能测试,主要分为正确性测试和执行速度测试。测试结果表明,本文所提出的并行实现方案和性能优化使得JPEG2000解压缩不仅能够正确实现,而且其速率远超过指标要求,可以满足实时解码的需求。从而,该系统可以应用到地面接收端进行实时图像解码处理。
[Abstract]:With the rapid growth of the research and development demand for Earth observation satellites, the amount of remote sensing image data to be processed increases rapidly, which poses a severe challenge to the development of the ground image detection and processing system. In view of the characteristics of high bit-rate, complex data processing and fast operation in the current data transmission channel, people pay more and more attention to the high-speed decoding of large amount of remote sensing images. The chip loaded on the satellite uses JPEG2000 compression algorithm. At present, the chip has been able to compress the captured images in real time, but the ground receiver has not been able to achieve real-time decompression. This paper focuses on solving this practical problem. In order to solve the problem of the complexity of JPEG2000 compression standard algorithm, this paper develops a high performance JPEG2000 real-time decoding processing system based on multi-core processor. Multi-core processors integrate multiple complete computing engines into a single processor, in which each engine can compute in parallel independently. Compared with the traditional single-core processor platform, the multi-core processor has a high performance. The multi-core processor platform used in this paper is composed of 4 TILE-GX36 processing cores produced by Tilera, each of which has 36 independent cores. In order to decompress images quickly and reliably, this paper embed the JPEG2000 decoding algorithm into the multi-core processor platform, and effectively optimized the decoding algorithm and the scheduling mechanism of the platform. Finally, the urgent problem of low real-time rate of decompression in the current task is solved. In the main part of this paper, the development history and present situation of JPEG2000 algorithm and multi-core processor are summarized, and the core technology of multi-core processor is briefly introduced. Then, the core algorithm of JPEG2000 coding and decoding is discussed in detail. In this paper, we propose and implement a high speed parallel JPEG2000 decoding scheme based on multi-core processor Tilera platform. The parallel scheme is not only between multiple processor cores, but also among 36 smaller execution cores on one processor kernel. In addition, the decoding algorithm and multi-core platform scheduling are further optimized, which greatly improves the decoding speed. Finally, the JPEG2000 decoding performance test based on Tilera platform is carried out, which is divided into correctness test and execution speed test. The test results show that the proposed parallel implementation scheme and performance optimization make the JPEG2000 decompression not only be implemented correctly, but also its speed far exceeds the requirements of the index, which can meet the demand of real-time decoding. Thus, the system can be applied to real-time image decoding at the ground receiver.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN919.81

【相似文献】