A Study And Implementation Of JPEG2000 High Speed Decoding Based On Many-core

JPEG2000 is a new generation of image compression encoding standard which uses the discrete wavelet transform as the core of the algorithm, the EBCOT as the core encoding algorithm. Compared with JPEG, JPEG2000 has a better low bit compression performance, support encoding image quality, resolution, color components, spatial position, and so on, a variety of progressive transmission order, as well as supporting the region of interested regions encoding and so on. JPEG2000 is widely used in military, remote sensing, medical, meteorology, and digital cinema because of its excellent encoding performance. At present, the research platform of JPEG2000 encoding and decoding is mainly CPU, DSP, FPGA and GPU, and most of the JPEG2000 decoding system has some limitations in the decoding performance, development difficulty and application range. Therefore, to design and implement a JPEG2000 high speed decoding software which is provided with high decoding performance, short development cycle, low cost of development and wide application areas, has important research significance and application value.This paper takes many-core CPU as the research platform, studies the implementation of JPEG2000 high speed parallel decoding, and on this basis, it is designed a high-speed parallel decoding of JPEG2000 system based on many-core. The main contents and results are as follows:1. Design of the high-speed parallel decoding system based on many-core platform. The multi thread technology is adopted in this paper, and Created the N sub thread, each thread is responsible for a encoding image decoding operation, decoding process is independent of each other. These independent parallel decoding channels consist of the main thread and thread forms the N sub, there by greatly improving the degree of parallelism and decoding speed.2. In order to reduce development effort and development time, this paper compared the decoding performance of Kakadu, Open JPEG and Jasper three JPEG2000 open source software, and based on the best decoding software, analysis its core structure, to prepare for the subsequent optimization.3. In order to meet the high speed requirement of encoding image, the input and output interfaces of the decoding program are optimized for the buffer input and output interface. For the input interface, this paper implements the J2 K and JP2 two format image data cache input by modifying the membership functions and the member variables of the virtual base class of the input interface. Accordingly, for the output interface, the output cache of BMP and RAW are implemented. In order to facilitate subsequent data operations, the decoded image data is stored in the corresponding data structure.4. In order to verify the performance and decoded image quality decoding software, this paper first researched the parallel high-speed storage technology based on many-core platform. Secondly, for the problem of multi-threaded operation can improve the decoding speed but will reduce the problem of disk storage speed. This paper designed a multi-channel high speed parallel storage mechanism based on memory queue on the many-core platform.In order to test and verify the performance of the JPEG2000 high speed decoding system based on many core, this paper conducted extensive testing experiments. Firstly, this paper tested the performance of the decoding system, which verified that the decoding system can greatly enhance the decoding efficiency. Secondly, in order to fully test the performance of the decoding system, the system was transplanted to many-core CPU platform. This paper tested the effects of different resolution images, different color component, different bit depth and different compression ratio on the performance of the decoding system. Finally, this paper tested to verify the decoding quality of JPEG2000 speed decoding system. The resulting analysis of experimental data shows that the design of the JPEG2000 parallel decoding high-speed system based on many-core CPU, has very high decoding speed and image reconstruction quality for 2x-16x times of the compressed encoding images.