Content-based processing and analysis of endoscopic images and videos: A survey
The main question we try to answer in this work is whether it is feasible to employ super-resolution (SR) algorithms to increase the spatial resolution of endoscopic high-definition (HD) images in order to reveal new details which may have got lost due to the limited endoscope magnification inherent to the HD endoscope used (e.g. mucosal structures). For this purpose we propose a SR algorithm, which is based on the Projection onto convex sets (POCS) approach. This algorithm is able to avoid over-sharpening, which is often seen with other methods. Since POCS-based approaches are iterative ones, we also propose an adaptive iteration scheme. We compare the quality of the reconstruction of our method against the quality achieved by other SR methods. This is done on standard test images as well as on images obtained from endoscopic video frames. We show that, while our approach produces competitive results on standard test images, we are not able to reveal new details in endoscopic images for various reasons.