Research On Infrared Super-resolution Imaging Technology Based On Micro-scanning

With the development of science and technology, infrared high resolution images are increasingly urgent required by military, medicine and many other fields. However, the way by improving detector to promote the resolution of images is not only costly but also limited by current manufacturing process. Using complementary information from multi-frame images, the super resolution imaging technology can improve the spatial resolution of images without changing the hardware condition of devices. Furthermore, the micro-scanning can collect multi-images with sub-pixel shift in same scene, thus optimizing the final imaging quality. Therefore, it is of great significance to carry out the research on infrared super-resolution imaging technology based on micro-scanning in the infrared field.This thesis is originated from the national natural science foundation of China, and focuses on the research of micro-scanning, image registration, super-resolution image reconstruction and other related technologies. Firstly, the thesis uses piezoelectric ceramics as micro-scanning actuator to design a multi-frame infrared images acquisition system and works out two kinds of super-resolution imaging schemes which respectively correspond to the non-uniform and uniform micro-scanning modes according to the system. Secondly, in view of the non-uniform micro-scanning, the thesis takes the high frequency information pyramid which combines the Lucas&Kanade optical flow algorithm instead of Gauss pyramid to achieve sub-pixel registration. And a total variation regularization super-resolution reconstruction method based on Retinex theory is proposed to reconstruct the images after registration. The method makes it possible to select regularization parameter adaptively according to image properties without artificial experience, which achieves the goal of improving efficiency and preserving the edge and the detail information of the infrared images. Finally, in view of the uniform micro-scanning, a quasi-curvature interpolation method based on the lateral inhibition mechanism is proposed. The method combining adjusted lateral inhibition model based on bimodal Gaussian distribution and reduced iterative curvature based interpolation is superior to the iterative curvature based interpolation in terms of the reconstruction result and the running speed.The thesis utilizes infrared images collected by infrared thermal imaging system in our laboratory to conduct the simulation experiments of above research. The experimental results show that multi-frame infrared image with sub-pixel shift can be collected by micro-scanning method based on piezoelectric ceramics. The image registration method corresponding to the non-uniform micro-scanning mode can achieve a displacement vector estimation of sub-pixel accuracy with a shorter operating time, and the total variation regularization super-resolution reconstruction method based on Retinex theory can acquire high quality infrared image whose mean square error, peak signal to noise ratio are improved many times, and structural similarity index is much higher compared with traditional method. Moreover, the quasi-curvature interpolation method based on the lateral inhibition mechanism can acquire reconstruction images with great visual effect and excellent real-time performance.