Research And Implementation Of DWT In JPEG2000

With the ever-growing multimedia technique, image compression requires higher performance as well as new features such as progressive transmission, region of interest and embedded stream etc. To meet such need in the special area, a new standard of still images-JPEG2000 is made out.Because of its high complexity of computation and the large requirement for memory, so its realization is strict with the hardware platform, normally the common embedded CPU can't meet its requirement while DSP and FPGA can perform well. This thesis will emphasis on DWT in JPEG2000 encoding algorithm and the way to implement it.Firstly, in the dissertation, dominance of JPEG2000, its mainstream of realization and the development of its realization is presented, then we research on the theory of wavelet transform and the main parts of JPEG2000 encoding algorithms.Secondly, according to the characteristic of its algorithm, we select ADI Company's ADSP TS201S chip as our encoding chip, and the algorithm is programmed in TS201 assembly language. According to the feature of TS201's hardware structure, we use a structure of the encoding process which is different from the structure of the realization on PC. The core modules of JPEG2000 standard--wavelet transform'optimization on data transfers, code parallel and memory allocation is described in detail. We provide three test images to evaluate the performance. The result makes it clear that the JPEG2000 software got the pretty compression performance and execution performance.The end of this article, because of getting stronger parallelism by designing of JPEG2000 encoding based on FPGA, then the system could get real-time requirement, the methods of the realization of wavelet transform based on the FPGA is also studied, and we combine a pipeline 1D-DWT and a line-based 2D-DWT together, then a new method of the realization of wavelet transform is given, which can reduce the system memory requirements, and improve encoding efficiency.