通信专业文献及翻译基于JPEG技术融合DSP和FPGA的可纠错图像传输系统.docx

1、通信专业文献及翻译基于JPEG技术融合DSP和FPGA的可纠错图像传输系统英文原文Mixed DSP /FPGA implementation of an error-resilient image transmission system based on JPEG2000Marco Grangetto,Enrico Magli, Maurizio Martina, Fabrizio VaccaAbstract This paper describes a demonstrator of an error-resilient image communication system over wi

2、reless packet networks, based on the novel JPEG2000 standard. In particular, the decoder implementation is addressed, which is the most critical task in terms of complexity and power consumption, in view of use on a wireless portable terminal for cellular applications. The system implementation is b

3、ased on a mixed DSP/FPGA architecture, which allows to parallelize some computational tasks, thus leading to efficient system operation. 1 Introduction Nowadays, there is a growing interest in the end-to-end transmission of images, especially motivated by the short-term deployment of next generation

4、 mobile communication services (UMTS-IMT2000). However, transmission in a networked, tetherless environment provides both opportunities and challenges. The wireless context implies that the data may undergo bit errors and packet losses, making it necessary to foresee error recovery modalities. It is

5、 thereby necessary that image communication techniques are provided with the ability to recover, or at least conceal, the effect of such losses. The forthcoming JPEG2000 image compression standard has been designed to match these requirements, and embeds some error detection and concealment tools. T

6、his paper addresses the development of a demonstrator of an error-resilient JPEG2000 decoder implementation for image communication over a lossy packet network. The robustness to packet erasures is achieved by combining the flexibility of the JPEG2000 framework with the powerfulness of source-channe

7、l adaptive, optimized Reed-Solomon codes. The decoder implementation is particularly significant in the context of wireless portable terminals for next-generation cellular systems, where the limited power budget and available dimensions impose severe constraints on the design of a multimedia process

8、ing system. 2 System overview In the following we provide a brief description of the functional units of the implemented system. 2.1 JPEG2000 image compression JPEG2000 is the novel ISO standard for still image coding, and is intended to provide innovative solutions according to the new trends in mu

9、ltimedia technologies. At the time of this writing, the standard is in advanced publication stage; the Final Committee Draft is the most recent JPEG2000 description publicly available, which our implementation conforms to. JPEG2000 not only yields superior performance with respect to existing standa

10、rds in terms of compression capability and subjective quality, but also numerous additional functionalities, such as loss less and lossy compression, progressive transmission, and error resilience. The architecture of the JPEG2000 is based on the transform coding approach. An image may be divided in

11、to several sub-images (tiles), to reduce memory and computing requirements. A biorthogonal discrete wavelet transform (DWT) is first applied to each tile, whose output is a series of versions of the tile at different resolution levels (subbands); then, the transform coefficients are quantized, indep

12、endently for each subband, with an embedded dead-zone quantizer. Each subband of the wavelet decomposition is divided into rectangular blocks (code-blocks), which are in- dependently encoded with EBCOT (Embedded Block Coding with Optimized Truncation) ; this latter is based on a bit-plane approach (

13、i.e. the most significant bits of the subband coefficients are transmitted first), context modeling and arithmetic coding. The bit stream output by EBCOT is organized by the rate allocator into a sequence of layers, each layer containing contributions from each code-block; the truncation points asso

14、ciated with each layer are optimized in the rate distortion sense. The final JPEG2000 bit stream consists of a main header, followed by one or more sections corresponding to individual tiles. Each tile comprises a tile header and a layered representation of the included code-blocks, organized into p

15、ackets. In order to form a progressive bitstream, i.e. one that can be only partially decoded with minimal penalty, the layers are formed and ordered in such a way that the most important information is placed at the beginning of the bitstream. The JPEG2000 decoder performs exactly the same steps (e

16、xcept for rate allocation), in reverse order: syntax parsing, codeblock decoding by EBCOT, inverse quantization, inverse DWT, and tile mosaicking; this is sketched in the right-hand-side box of Fig. 1. 2.2 Adaptive Reed-Solomon packet protection Although JPEG2000 embodies advanced error concealment

17、techniques to mitigate the effect of errors, it does neither contain, nor specify any error correction method, in order to recover lost packets. On the other hand, packet losses are likely to occur in a network potentially subject to congestion, as is often the case in practice. In order to overcome

18、 this problem, a technique has been recently proposed, called Unequal Loss Protection (ULP), and based on the joint use of RS codes and packet interleaving, as shown in the left-hand-side of Fig. 1. Let us consider a maximum rate allocated to the image transmission, e.g. N packets of size L; the sou

19、rce bitstream is rowwise inserted in the interleaving matrix, followed by a proper amount of parity symbols, say Ti for the i-th row. The packets are read on the columns of the interleaver. The allocation problem consists in finding the optimal partitioning between source and code symbols for each r

20、ow of the interleaver, so as to maximize the quality of service at the receiver; see for implementation details. At the decoder, due to the error correction capability of RS codes, the i-th row can be exactly recovered provided that the number of packet erasures has been less than Ti. Figure 1 Syste

21、m architecture 2.3 Proposed system In this paper we present a demonstrator of a complete decoder for image communication over a lossy packet network. The decoder consists of 1) ULP decoder, followed by 2) JPEG2000 decoder. The goal is to demonstrate that such tasks can be effectively performed on a

22、modern DSP, satisfying real-time operation requirements. This objective has been accomplished by using a mixed DSP /FPGA architecture. In particular, the implementation has been developed using a Texas Instruments (TI) TMS320C6201 DSP board as target device. The additional bitstream protection proto

23、col, based on RS codes, has been implemented on a Virtex XCVl000 FPGA device from XILINX, which is able to concurrently exchange data with the DSP. As is shown in Sect. 4, the developed system is able to perform very fast decoding of still images, at a rate compatible with real-time video applicatio

24、ns; thus, an extension to MotionJPEG2000 video coding can be envisaged, with minor modifications of the current demonstrator. 3 System architecture Resorting to mixed DSP /FPGA based architectures allows to achieve very-high performance systems, with excellent properties of reconfigurability. Prelim

25、inary partitioning studies have put into evidence that wavelet transform and EBCOT are the most demanding tasks in JPEG2000 decoding process. However, since many extensions of the standard are still possible, an all-DSP implementation offers an excellent degree of reconfigurability. On the other han

26、d, FPGA is well suited to the implementation of a Reed Solomon core, needed to grant error resilience of the communication system. It is worth noticing that, with an ULP, the RS decoder needs to be re-adapted to work with words of bits which can keep different amounts of information symbols and, as

27、a consequence, different amounts of protection symbols. The need for changing these parameters on-the-fly is perfectly fulfilled by an FPGA, by simply loading new values into some configuration registers. Moreover, the high memory bandwidth needed to quickly de-packetize and de- code the received bi

28、t stream makes the use of a Reed Solomon FPGA implementation very attractive. The JPEG2000 decoder module, entirely implemented on DSP, is composed by four main blocks: syntax parser, entropy decoder (EBCOT), uniform scalar dequantizer, and inverse wavelet transform. Moreover, two additional tasks,

29、devoted to communication management between DSP, FPGA and a personal computer, have been introduced. 3.1 Syntax parser The parser is the functional block that interfaces the JPEG2000 decoder with the RS decoder. It retrieves RS decoded packets, and extracts from the compressed JPEG2000 bitstream all

30、 the relevant information needed to perform image reconstruction. Firstly, the bit stream main header is read, which contains information on the parameters used during the encoding process (e.g. image size, wavelet filter used, number of decomposition levels, quantization thresholds, and so on). Aft

31、er that, tile headers are read, which provide information specific to each image tile. Finally, each packet contained in the bit stream is read, and the data and parameters of each codeblock are extracted, and fed as inputs to the EBCOT decoder. 3.2 EBCOT Right after the bitstream syntax parser, the

32、 subsequent stage in the JPEG2000 decompression chain is the entropy decoder (EBCOT). From an algorithmic point of view, EBCOT is a block-based bitplane encoder followed by a reduced complexity arithmetic coder (MQ). It subdivides each wavelet subband into a disjoint set of rectangular blocks, calle

33、d code-blocks. Then the compression algorithm is independently applied to every code-block. The samples of every codeblock are arranged into so-called bitplanes. To decode a code-block, EBCOT always starts from the most significant bitplanes, and then moves towards the least significant ones. The compressed information of every code-block is th