Physically Motivated Reconstruction of Fiberscopic Images

Flexible endoscopes are applied in modern techniques for technical inspection as well as medical diagnostic and therapy in keyhole-surgery-scenarios. Their characteristic bendable image conductor consists of a limited number of coated fibers. An optical fiber consists of a core surrounded by a cladding layer. This configuration leads to imaging artifacts, called comb structures. They have a negative impact on further image processing steps, like feature detection and tracking. The intensity distribution of a cross-section of a fiber is usually modeled by a two-dimensional Gaussian. We propose a preprocessing algorithm which exploits this physical property to remove the comb structure while retaining the image content. The proposed approach effectively removes of fiberscopic comb structure in real-time. It is adaptive to arbitrary endoscope and sensor combinations. The results give prospect of a large field of possible applications both for visual optimization in the clinical environment and for further digital imaging tasks.