Design And Implementation Of Quantization For On Board Wavelet Image Compression System

With the development of satellite remote sensing technology, the high-resolution remote sensing satellite images are widely used to all fields of the national defense. Every country is trying to develop high-resolution sataellites for military uses. However, images acquired on board spacecrafts (earth observation satellites, scientific probes, etc.) represent in most cases very large volumes of data. It is then necessary to store these data on board and to transmit them to the ground. Due to the stringent limitations (in terms of mass, power and cost) which apply to on board equipments, it is very essential to reduce to a minimum the on board storage capacity and the on board transmission rate needed to fulfill the mission. There exist various compression algorithms that provide good data reconstruction performance at high data reduction, most notably the International Organization for Standardization (ISO) JPEG2000; however, they are very expensive, not appropriate for real-time implementation, and none has adequately addressed the implementation requirement arising from push-broom instruments for space applications. To overcome these limits, we have made deep research on remote sensing images with "Satellite video compression"as the background, and a novel and efficient adaptive quantization method for on board remote sensing image compression based on DWT is proposed. For DWT-based compression, the efficiency of a compression scheme highly depends on quantization process. Our method is based on an accurate modelisation of the rate-distortion curve. It provides optimal dynamic bit allocation for sub-bands of DWT and realizes adaptive quantization for wavelet coefficients. At last, it provides exact rate control for the coder. The experiments show that the efficiency of this method is very close to JPEG2000 with a very low complexity. Furthermore, it is well-suited for parallel systems in the sense that sub-bands of DWT are quantized and encoded separately. The FPGA design of a high speed dynamic quantizer for DWT-based compression is implemented in this thesis. The quantizer can perform dynamic quantization for wavelet coefficients after the determination of the best quantization steps {qi} that minimize the average mean square error at a given bit rate. The signed divider, which is the crucial part of the quantizer, is implemented using an IP core that adopts RTL Verilog HDL to design. This quantizer has achieved high timing performance with a fmax at 226.30MHz. In addition, the designs of a dynamic inverse quantizer and a dynamic scanning module for quantization symbols are given. All designs have been synthesized and simulated successfully on the platform of Quartus II 4.0. The results of simulation and FPGA validation show that the designs have met the requirements of function and timing and been provided with robust performance.