The Research On A Novel Super-resolution Reconstruction Method

High resolved imaging technologies and even super-resolved ones are always the hot topics in the electronic-optical domain. As optical imaging system transits to Electro-Optical imaging one, a new question appears that the resolution is mainly limited by detector. It is obvious that the prior is to enhance detector resolution for the high-resolved or even super-resolved imaging system. Furthermore, the main influencing factor limiting detector resolution is the spectral aliasing by the undersampling phenomenon, so super-resolution reconstruction is usually used to improve the imaging quality.In terms of the problem that how to enhance the detector resolution and make the system resolution not being affected by it, we take a lot of researches on the E-O imaging performance analyses and classical super-resolution reconstruction methods. Then we attempt to use some unconventional imaging theories and technologies to break some limits existing in recent methods. During the period, we get some meaningful theoretical-and-experimental results, as the followings:1. Three types of super-resolution reconstruction methods are summarized that microscanning for focal plane arrays, sub-pixel technology for linear arrays and spectral coding without movement elements.2. In theory, super-resolution reconstruction is proved available from both views of MTF modeling and sensitivity. Three main influencing factors of E-O imaging system are analyzed that diffractive cutoff, detector cutoff depending on the size and shape of active area, and Nyquist frequency which is the highest frequency reconstructed without aliasing. Through computing the relation among them, we make sure that the resolution is related to the size of active area with original detectors by super-resolution reconstruction. What's more, Overall system-resolution is limited by detector for the low-filterring effect of active area.3. Considering it depends on the size and shape of active area as well as the sampling process, spatial resolution is enhanced further unless the cutoffs between optical system and detector match better after obtaining sufficient sampling frequency. Focal-plane coding method is proposed to realize this idea. With two detector-arrays of special-shaped pixel where each pixel is deducted a quarter, multiple mis-registered frames are got to solve the grayscale matrix and reconstruct super-resolution image in the post-detector processing. We present the theoretical assessment that if a small plot of active area per pixel is deducted and the frequency-response distribution of the complement will equal to that of the deducted part without consideration of the amplitude.4. We prove the idea available from both theoretical and experimental views. We set up point spread function of the special-shaped active area and the equations imply the method should increase detector cutoff frequency two times. Furthermore, we also take a MWIR experiment. We choose initially two 46μm×46μm linear arrays which are 100% full factor and cut down a quarter each pixel. The result demonstrates the bar-contrast interval is discriminated in objective space up to 3.1mm after optical system composed of a 6000-mm f/10 collimator and a 50-mm IR lens.