| Passive millimeter wave (PMMW) imaging system forms images by taking advantage of different radiation characteristic between targets and background. It has a great value of application in the fields of military reconnaissance, anti-terrorism, maintenance of stability, landing of aeroplane and so on. However, current system does not have ideal resolution to meet our requirement. Multi-frame superresolution technology has the ability to enhance image resolution without any change of hardware structure, which has become a popular cutting-edge technology with important application prospect in both military and civilian fields. Supported by National Natural Science Foundation of China, passive millimeter wave image multi-frame superresolution reconstruction technology has been deeply investigated in this dissertation, the main contents are as follows:1. The operating principle of passive millimeter wave imaging system and image degradation process have been analysed. On the basis of subpixel analysis technology, mathematical model of passive millimeter wave multi-frame imaging system and the non-uniform sampling structure in high resolution grid formed by projecting low resolution sampling points on it are provided.2. To solve the limitation that superresolution algorithms based on statistical optimization can only be applied to single-frame image, under the assumption of independent identically distribution, improve these method applicable to multi-frame situation by using single pixel method to derive conditional probability density function. These methods can accomplish spectrum aliasing eliminating and high frequence extrapolation simultaneously.3. Considering the problem that the value of peak point and its nearby area are susceptible to noise in traditional Extend Phase Correlation (EPC) method, resulting in inaccurate displacement parameter, use Phase Correlation Partial Resampling (PCPR) algorithm based on IDFT to compute subpixel level motion parameter. This algorithm significantly improves the accuracy of subpixel level motion parameter in different background noise. 4. Analyzing the importance of regularization parameter, which determines the effectiveness of computed scale and wavelet coefficients for high resolution image in MBFR algorithm, an Iterative Computation of Regularization Parameter algorithm (ICRP) for scale and wavelet coefficients of passive millimeter wave high resolution image is researched. A better reconstruction result can be achieved by applying this algorithm.5. To solve the problem that some superior prior information is failed to use effectively in wavelet domain, a Multi-frame Conjugate Gradient super-resolution reconstruction algorithm in Wavelet-domain (MCGW) is researched based on Generalized Gaussian Distribution and advantage of conjugate gradient method in convergence rate. Combined with ICRP, this algotithm can not only enhance resolution of images, but also expand range of extrapolated spetrum.Experimental results verifies that multi-frame superresolution algorithms based on statistical optimization and wavelet analysis realize fusion of subpixel information provided by image sequcence, accomplish high resolution image reconstruction and spectrum extrapolating, thus enhance resolution of images captured by passive millimeter wave imaging system. |
PDF Full Text Request