Realization And Optimization Of Wavelet Transform In JPEG2000 System On DSP-Based

As a new generation of static image compression coding standard, JPEG2000 abandoned the traditional block coding method, which mainly use the DCT, by replacing it with a wavelet transform-based multi-resolution coding algorithm as a core algorithm. It marks the wavelet coding will become the mainstream technology of images and video encoding. In the field of image coding, high-performance and high-precision is one of the objectives pursued by the people. Therefore the use of high-performance digital signal processor (DSP) for efficient implementation of the float wavelet transform customized by JPEG2000 has practical significance. This thesis firstly expounds some of the basic concepts of the wavelet transform theory, followed by the core structure of the DSP, which is selected, as well as the ideas of optimization associated with it. Then this thesis presents a multi-level wavelet transform method which is completely completed in the on-chip memory of the DSP, to eliminate the impact of a level of wavelet transform on the amount of data exchanged between on-chip memory and external memory, and reasonability arranges the pipeline, to complete multi-line wavelet transform of the image data at the same time. The experimental results show that the method guarantees high precision and improvement of the wavelet transform data throughput. The system has been implemented on DSP.