Research On Super-resolution Reconstruction Method Of Passive Millimeter Wave Image

Passive millimeter wave imaging is a practically valuable probe technology. Due tothe inherent characteristics of passive millimeter wave imaging, radiation image is alwaysin low resolution and lack of readability. Using hardware methods to improve the qualityof radiation imaging, we will face the problem of economic costs and technical difficulty.Image super-resolution is a technology to improve the resolution of imaging systems byupper software methods. Combined with the characteristics of passive millimeter waveimaging, applying the technology to radiation image is provided with practical value.Based on the basic principles of image super-resolution and radiation imaging, andthe single image degradation model of passive millimeter wave imaging, the process ofsuper-resolution recovery consists of two steps: reconstruction methods and learningmethods. Airspace method is the mainstream of image reconstruction, simulations andcomparisons demonstrated that the adaptive regularization method is more robust;“Example-based Super-resolution” was chosen to be the right learning algorithm for itsbroad applicability. Dynamic adaptive control progress of regularization method wasachieved by local characteristics of images. Train set and searching method of learningalgorithm was optimized to debase the example image dependence and improve thesearching veracity. In the transplant process of the algorithm, the technical problems ofestimate of point spread function and selection of training reference images were solved.Simulation and experimental results demonstrated that the super-resolution recoveryprocess is robust and effective, able to enhance definition and visual quality of millimeterwave radiation images.