Multi-observer survey on quality enhancement techniques for still image acquisition in colonoscopy
Sebastian Groß, Martin Schink, Thomas Stehlé, Axel Behrens · RWTH Publications (RWTH Aachen) · 2010
Colonoscopy, endoscopic examination of the colon, is a crucial part of the diagnostic and therapeutical process recommended for various diseases of the colon.During examination, still images are frequently taken for documentary purposes and later computer-based analysis.Still images without blur caused by movement and focus variation are difficult to capture because of limited endoscope manoeuvring control and colon peristalsis.This problem can be engaged by capturing a video sequence rather than a single image.A final sharp frame can subsequently be selected from the sequence via sharpness metrics.Three sharpness measures were evaluated: Variance, Sum Modulus Difference, and Perceptual Sharpness Metric.Alternatively, an approach called Extended Depth of Field aims at combining a refined image from the sharpness information of multiple frames.In this paper, a multi observer survey comparing the different approaches is detailed.Analysis of the survey results yields that sharpness considerations paired with performance reasons make Sum Modulus Difference the preferred algorithm for the selection of sharp images from colonoscopic video streams.