Study on the Image Processing Algorithms for Optical Readout Based Un-cooled MEMS IR Imager Systems

摘要

MEMS thermal transducers offer a promising technological platform for un-cooled IR imaging. We report on the fabrication and performance of an optical readout based MEMS IR FPA based on the bimaterial microcantilever. For the IR images of objects obtained by these FPAs, processed by the fast Curvelet transform denoising and inpainting algorithm, the image quality is improved obviously. Great compute and analysis have been realized by using the discussed algorithm to the simulated data. The experimental results demonstrate, better RMSE and highest Peak Signal-to-Noise Ratio (PSNR) compared with traditional methods can be obtained. At last we discuss the factors that determine the ultimate performance of the FPA. And we indicated that one of the unique advantages of the present approach is the scalability to larger imaging arrays.