Research Of Wavelet Denoising Algorithm Based On DSP For Digital Neutron Images

Digital neutron image, obtained from the digital neutron imaging system after neutrons penetrate through the sample, carries information about the sample's internal structure.Unavoidably affected byγ-ray interference, neutron scattering and CCD's electronic noise, neutron images always suffers from degradation, e.g. corrupted by intense noise, detail-blurred, thereby it is not advantageous to extract and analyze the information about the sample's internal structure.In the study of digital neutron image processing, the processing algorithms to deal with large amout of image data are so complex that the low-speed software-based procession can not meet the demand of realtime processing. With the birth and development of DSP, FPGA chips, people focus more on the hardware approach of image procession. Digital signal processor (DSP) has the characteristics of high-speed procession and single-cycle instruction providing complex functions; therefore, it becomes more and more popular in the applications of digital image processing. In particular, transplantation of wavelet image processing algorithms to the DSP hardware can improve processing speed, algorithm's confidentiality and real-time performance. It is a new trend of effective image processing.According to the above analysis, based on DSP hardware and PCI bus, a high-speed Digital neutron image processing system is designed to improve neutron image processing speed. The system uses DSP and PCI interface chips as a hardward platform. The DSP carries out algorithms, manages system's operations, and achieves the system auto boot. PCI interface chip is used for high-speed PCI bus interface from DSP to PC.On the DSP simulation platform, denoising algorithm for neutron images is studied. Contraposing the noise characteristics of neutron image, wavelet threshold de-noising methods are chosen. The paper achieves and improves the wavelets algorithms on the chips of TMS320VC5416 of TI Company, based on which the threshold algorithms suitable for fixed-point DSP are introduced.Results show that this system achieves high processing speed, meets the real-time requirements and has good confidentiality.