Research On EBC Algorithm Of JPEG2000 For Efficient DSP Implementation

Because of the limited wireless transmission capability, high fidelity and low bit rate image compression technology becomes one of the key technologies in the aerospace sensing area for real-time image transmission. Among all types of image compression algorithms, JPEG2000 is the state-of-the-art image compression standard which has many advantages, such as high compression efficiency and high image fidelity. However, because of its high computational complexity, real-time JPEG2000 hardware implementation becomes the biggest problem when using it in the aerospace sensing area. Supported by the military project"Development of×××optical imaging satellite"and the 863 National High Technology Research and Development Program"×××image data compression technology"(Grant No. 2006AA701121), this dissertation concentrates its attention on the Embedded Block Coding (EBC) algorithm in JPEG2000.The main contents and contributions of this dissertation are as follows:1. The characteristic and implementation flow of EBCOT algorithm are studied. The emphasis is put on the theory of EBC algorithm which is the EBCOT tier-1. Combined with the DSP characteristic, the shortcut of the EBC framework for DSP implementation is analyzed. Because of the multiple loops in the EBC framework and the many conditional operations in the program, DSP software pipeline can not be established if the existing EBC framework is used.2. Some existing improved EBC frameworks are analyzed. However, they are only aiming to improve the EBC algorithm itself. The hardware implementation of the EBC algorithm is not considered in the improvement. Aiming at the shortcut of the EBC framework in the DSP implementation, an improved EBC framework which is suitable for the establishment of DSP software pipeline is proposed in the paper. The arithmetic coding module is separated from the framework and the bit plane scanning is unrolled. Then the sign coding (SC) is put to implement before the bit plane scanning. After the improvement, the DSP pipeline of the scanning implementation can be established successfully which greatly faster implementation speed.3. Some optimization schemes are proposed base on the proposed EBC framework. The arithmetic coding module is improved for the high efficient parallel DSP implementation. And the DSP implementation method of SC is proposed. Then the storage structure of the wavelet coefficients is modified.4. Based on the improved EBC algorithm, the DSP program is designed. Some tests are done on the JPEG2000 hardware platform which consists of a FPGA and some parallel DSPs. It is shown that the proposed EBC algorithm accelerates the DSP implementation speed greatly.Now it is used in a JPEG2000 based remote sensing image compression equipment prototype.