Automatic glare removal in endoscopic imaging

Background: Glare from surgical instruments and tissue surfaces often occurs during endoscopic procedures and can be disturbing to the operator. The brightness level of the light source can be reduced, but at the expense of overall image clarity, so alternative solutions are needed for removing glare. Digital image-processing methods offer the opportunity to lessen or eliminate glare by reducing the intensity of the affected parts of the image. This study investigated a new automated method for glare reduction that uses two different intensity thresholds as a basis for applying glare reduction processes and it also reduces unpleasant artifacts at the glare region boundaries. Methods: The new glare-reduction method was compared with a previous method. Three variants of each method, each with different color biases in the glare regions, were applied to a 20-s surgical recording containing substantial amounts of glare. The six versions and the original recording were evaluated subjectively by a group of 10 experienced surgeons using a paired-comparisons method, in which each version was compared for preference with all the other versions. Results: The new double-threshold intensity-subtraction method scored significantly higher than the previously developed glare-reduction method (p <0.05). It also scored higher than the original (unprocessed) version, but not significantly. The color bias was important, with combinations of pink and grey performing better than yellow tints. Conclusions: The findings show the new method to be a significant improvement in automatic glare reduction compared with earlier methods. The method is not computationally demanding, so it can in the future be evaluated clinically in high-definition endoscopic imaging systems and developed further in this environment.