A Fiber-optic Endoscope Honeycomb Structure Removal Algorithm Based on Interpolation of Valid Neighborhood Pixels

Fiber-optic endoscopes consist of thousands of tiny optical fibers, and the honeycomb structure caused by the spacing between these fibers may lead to the degradation of the image quality. Existing honeycomb structure removal algorithms that are based on conventional filters or deep neural networks have shortcomings of image sharpness decrease or high calculation consumption. In this paper, a novel honeycomb structure removal algorithm based on the interpolation of effective pixel points in the neighborhood is proposed. Besides, a honeycomb structure generation algorithm is proposed to simulate honeycomb structure on different images and different algorithms were compared based on the simulated images and U.S. Air Force resolution target images (USAF). The PSNR and SSIM evaluation metrics are compared, and the experimental demonstrated that the proposed algorithm presents better removal results due to the interpolation based on effective neighborhood pixels other than all neighborhood pixels. The proposed honeycomb structure generation algorithm and the simulated dataset are available on GitHub: https://github.com/ll623/fiber-images.