Research On Map Based Super-resolution Method For Hyperspectral Imagery

Hyperspectral imagry(Hyperspectral imagery, HSI) has been applied in more and more fields. The spectral resolution of HSI is high, but its pixel resolution of the corresponding surface features is limited. A lower spatial resolution bring great difficulties to the data processing such as precise matching, mixed pixel separation, target detection, identification technology, etal. We can say that spatial resolution of hyperspectral has become a major constraint to imaging applications. Therefore, how to improve its spatial resolution catch more eyes in the academic world. Various academics are committed to improve the spatial resolution of the research, but have not yet get a good solution.Super-resolution reconstruction technique is a method which can remove and lower the reducing quality of the picture during the picture capture.The method recovers the picture by establishing a mathematic model and following the inverse process of the reducing picture's quality. It uses the low resolution resource to receive high resolution picture which has much value of research and application.In this paper, MAP-based super-resolution method, combined with high spectral characteristics of the HSI itself, establish a high-spectral image imaging model, and then super-resolution reconstruction technique has been applied to HSI to enhance its spatial resolution. The paper mainly focuses on the following aspects:First of all, it is described of the characteristics of HSI, of the imaging spectroscopy, as well as of the-state-of-the-art of improving HSI resolution to describes the background and application value of the research.Secondly, the extension researches are respectively done on classification, application, evaluation, mathematics foundations. The basic theory used in this study of MAP method is then highlighted as the theoretical foundation of the follow-up study.Next, hyperspectral image super-resolution model is proposed. Departing from the spectral haracteristic of HSI and the ordinary imaging methods, this article conclude the imaging model for high-spectral images. And in the establishment of the high-spectral imaging model, the original high-dimensional data is mapped to low-dimensional transform space in use of interesting categories, which greatly reduces the complexity of the algorithm and protects the interest categories.Finally, based on maximum a posteriori probability (MAP) theory and with the help of these high-spectral imaging models, super-resolution restoration method for high-spectral image is obtained. During the recovery operation, local analysis is used instead of global analysis to avoid huge computation burdun of super-resolution method coused by large-scale matrix operations. Through a variety of evaluation methods, and special applications-classification, simulation results show the verification method is effective and feasible.