Image Transmission through Incoherent Optical Fiber Bundle: Methods for Optimization and Image Quality Improvement

摘要

Artificial vision, in spite of all theoretical and practical progress accomplished during last thirty years, still cannot be employed technically in some hazardous and strong industrial areas, where conditions are such that cameras would not operate properly. Possible alternatives on such problem are the quite recent IP65 and IP67 industrial cameras, and associated connectors, protected by an anticorrosive, waterproof, and high temperatures resistive carcass employing a dedicated electronic, in addition robust to almost 3G accelerations. Nevertheless, such cameras are still very expensive compared to conventional industrial cameras and would still not be enough in hardest conditions (explosive gas or dust environment) or in electromagnetic interferences environments. A good alternative in extreme conditions is the use of optical fiber bundle. Since such cable only transmits light, they intrinsically have electromagnetic interferences immunity and, depending on the fiber material, could be exposed to very high temperature (over 1000℃ for sapphire fibers for example), could be employed in almost all corrosive environment, and are totally submersible. Nevertheless, the coherent optical fiber bundles are very expensive, and for large distances could be non-competitive facing the other hardware solutions (armor plating, electromagnetic interferences isolation). Thus, the best suitable option to develop a competitive system in those particular cases is the use of incoherent optical fiber bundle (IOFB), nowadays just employed for illumination tasks. Image transmission in this case is not direct but require a calibration step that we discuss in this paper. Improvement of the noisy resulting quality image is also exposed here, achieved by experimental post calibration methods. We propose in this work a new calibration method of incoherent optical fiber bundle (IOFB) for image transmission purpose. Firstly, we present the calibration method (an enhancement of previously published calibration methods), some resulting reconstructed images and a discussion on its quality improvement employing simple denoising methods assisted by low pass filters (smooth filter, resizing method, ...). We finally depict the two new post calibration methods, and its associated noise generator physical phenomena we want to treat. Those two post calibration method are: correction of the input plane optical aberrations and extraction of a unique pixel (almost centered) for each fibers. We finally present some resulting images that demonstrate how it efficiently refine the reconstruction Look Up Table (LUT) used for the output image reconstruction improving image quality, image contrast but also reconstruction processing time.