Super-resolution Reconstruction Algorithm Of Passive Millimeter Wave Image Sequence

The passive millimeter wave(PMMW) imaging has a stronger penetration for detecting the concealed dangerous goods compared to the infrared and ordinary microwave imaging. What’s more, PMMW imaging has no radiation damage to people which meets the security requirements. While, the wavelength of millimeter wave is longer than visible wave’s and the hardware limitations of imaging system cause that the PMMW imaging systems has severe diffraction limitation so that the direct obtained imaging image has low-resolution and do not meet the actual application requirements. Therefore, we must study the super-resolution reconstruction technology of PMMW image sequence to improve the system resolution.How to extract the accurate information between frames and to reconstruct the missing information exactly are not only the key points of super-resolution reconstruction of PMMW image sequence but also the hotspot in domestic and foreign researches. In this paper, the super-resolution reconstruction technology of PMMW image sequence has been research deeply and the main content includes the follows.1. After analyzing the basic principles of PMMW imaging technology, we obtain a mathematical model of super-resolution reconstruction of PMMW image sequence by introducing the sub-pixel displacement;2. For the image motion blur caused by the random jitter of the imaging system, we adopt the image motion blur judgment and recovery algorithms to achieve accurate restoration of the motion-blurred image. While for the motion blur angle distortion for PMMW images, the correction algorithm based on the non-uniform quantization is proposed to make the estimation error under 1° and ensure motion-blurred image restoration accuratly;3. The image feature extraction and feature registration algorithm is studied in this thesis. For the image registration sub-pixel accuracy is difficult to improve because of the affection of noise, we introduce block registration and clustering process to improve the accuracy and stability of the registration. The simulated and real data demonstrate the accuracy of the registration algorithm can achieve 0.05-pixel;4. For the ill-posed problem of regularization term in iterative solving the maximum a posteriori probability(MAP) based image sequence super-resolution reconstruction, we study an improved adaptive regularization MAP based image sequence super-resolution reconstruction algorithm by optimizing the regularization term to improve the noise filtering and edge reconstruction performance in the iterative solving process. Then we has optimized the solving process by using multi-grid processing algorithm which has better edge retention and faster computing speed compared with the conventional MAP algorithm;5. For the problem that the prior information cannot be introduced to the adaptive Wiener filtering(AWF) based super-resolution reconstruction algorithm, we established an improved AWF based PMMW image sequence super-resolution reconstruction algorithm through introducing the point spread function model of PMMW imaging system, so that the algorithm can adapt the characteristics of PMMW image better. Then, we optimizing the reconstruction process by using sliding window to enable the super-resolution algorithm can be used in a continuous imaging situation.