Study On Key Algorithm For Infrared Image Processing

Infrared imaging technique is becoming more and more usual, and expected to be used in many fields of the national economy, especially in the military field. However, infrared imaging system is limited by devices themselves and imaging mechanism, which makes infrared image poor and the performance advantages of infrared imaging technique restrained. Thus, at the same time of developing high performance infrared imaging system, the research of relative infrared-image process technique is also very important.This thesis has a deep investigate to infrared imaging and relative image processing technique, mainly aiming at Infrared Focal Plane Array (IRFPA) imaging system. Three important infrared-image process algorithms are put forward, for nonuniformity correction (NUC), image enhancement and object edge detection.Firstly, for the inherent nonuniformity in IRFPA and the influence that the temporal excursion of the detector response parameters and the nonlinear response imposed on the NUC, a Kalman-filter NUC algorithm based on the piecewise linear model is proposed. In this algorithm, a state equation and a measure equation are established for every detector in IRFPA, and then the recursion of the NUC based on Kalman-filter is deduced. On this basis, an extended Kalman-filter NUC algorithm by utilizing the piecewise linear model of detector response is developed to overcome effectively the effect that the nonlinear response of IRFPA detectors have on the NUC accuracy. Experimental results indicate that the improved algorithm acquires good effect of the nonuniformity correction. It not only inherits the advantage of the original algorithm that resolves the problem of the temporal drift in the gain and the bias in each detector by updating NUC parameters with information of the current scene, but also reduces the influence of the detector nonlinear response to the NUC performance, so it is more suitable for IRFPA under large response-range.Secondly, due to the low contrast and low signal-to-noise ratio (SNR) of infrared images, an infrared-image edge enhancement algorithm based on multi-scale morphological wavelet transform is proposed, which combines wavelet analysis and mathematical morphology. At first, the morphological wavelet transform is adopted to decompose the input infrared image, which extracts the characteristic of multi-scale edge of the image. Then a nonlinear enhancement operator is used to enhance the details of image edge under different scale. Finally, the inverse transform of the morphological wavelet is applied to synthesis image. The algorithm can avoid over-enhancement of noise and enhances contrast of image. Two groups of experimental results with difference real infrared-image demonstrate that the presented algorithm can enhance the infrared image edge effectively.Finally, for achieving effectively the edge-detection and the segmentation to the target, an infrared-image edge enhancement and segmentation algorithm combining lifting scheme wavelet transform and Snake model is proposed. In this algorithm, the lifting scheme wavelet transform is adopted to enhance infrared image and improve the contrast, then the Snake model is utilized to extract target edge information and accomplish the image segmentation. The experimental results with two groups of real infrared-image validate the presented method can achieve improved image contrast and extract more precise target edge.Nonuniformity correction, image edge enhancement and target segmentation are the key image-processing techniques that must to be solved that in the infrared imaging and its application system. These relevant image-processing techniques with high performance and adaptive capacity should be developed further by utilizing the modern signal-processing theory in the future.