Research On Multi-frame Super-resolution Algorithms Of Passive Millimeter Wave Based On The Reconstruction Method

Passive millimeter wave (PMMW) imaging is a kind of technology that theimaging system passively receives the energy of the natural radiation millimeter-wavethrough the scenes and objects，and it makes use of the different millimeter-wave energyintensity between the different scenes to image. The passive millimeter wave imaginghas its unique technical advantages compared with the infrared imaging and visible lightimaging and the PMMW has important and broad application prospects. The passivemillimeter wave imaging system can be used for detecting and imaging in a variety ofweather conditions throughout the day. The system has improved its dependence on thetime and climate when it is used for detection. The system itself does not radiateelectromagnetic waves so that it can break many limits for the application. Theadvantages of the technology make it widely used in military and civilian areas such asthe scene for monitoring, the safety inspection and the military reconnaissance.However, the characteristics of the system also make itself get some technicalshortcomings. For example, the constraints on the antenna effect of the imaging systemand its hardware implementation leads to the lower resolution of the image which cannot completely satisfy the above application needs.Multi-frame super-resolution algorithm will take advantage of the signalprocessing methods to ameliorate the quality of the millimeter wave image withoutchanging the condition of the system hardware. This paper relied on the nationalpre-research projects do deeply research on algorithms about multi-framesuper-resolution in PMMW The main work and conclusion are as follows:1Through the analysis on passive millimeter wave imaging system and its imagingprocess, a mathematical model of the image degradation is being studied. Themulti-frame image super-resolution algorithm model is introduced on the base ofsingle-frame super-resolution algorithm model by the introduction of sub-pixel celltechnology.2To solve the problem that the estimation of the sub-pixel motion between thepassive millimeter wave multi-frames, the digital image motion model is being discussed at first and the concept of the “bad frame excluding” is being proposed. Thenblock-matching sub-pixel estimation method and the optical flow sub-pixel estimationmethod has been studied. At last the simulation experiments of the sub-pixel motionestimation method based on optical flow has been done to make sure the estimation ofthe method can meet the needs of the coming algorithms mentioned in the next twochapters.3To solve the problem of super-resolution algorithm based on projections ontoconvex sets (POCS), the image amplification method based on edge-directedinterpolation is used to build the initial estimate of the high-resolution which has theminimum error. Then this paper shows the analysis of the characteristics of the gaussianmodel of the point spread function and tries to use the bilateral filter to get instead of thegaussian model PSF so that the spatial domain of the image can contact with the pixeldomain. In the next, the deficiencies of the template transition point spread function arerectified. The high-resolution millimeter wave image is reconstructed in the end.Compared with bilinear interpolation and super-resolution algorithm based on POCSalgorithm, the high resolution image is achieved by the improved algorithm4Analsizing the based on regular multi-frame super-resolution algorithm, wepropose a regularization method based on improved passive millimeter wave imagesequence reconstruction. First, the interpolation method based on partial differential(PDE) for image is used in the algorithm and the improved Laplacian operator is alsoused to characterize the priori information of the image. Then the adaptiveregularization factor is selected for the reconstruction and the high-resolution millimeterwave image is reconstructed in the final.. The algorithm can obtain a betterreconstruction results because it can extrapolate the high frequency information of theimage, and effectively improve the image resolution.The result of the experiments with the simulation image and measured passivemillimeter image show that the algorithms can make use of the sub-pixel informationbetween the multi-frame images to extrapolate the high frequency information of theimage so that the image resolution. can be enhanced effectively.