A Colour Rotation Based Detector for Bleeding Gastric Pathologies

Abstract

Colour is one the distinct features in computer vision systems and can be represented in several colour spaces. The most commonly used is the RGB colour space, since most high-quality cameras used in computer vision systems provide RGB signals, for example endoscopic capsule cameras. This work describes a computer vision algorithm intended to assist the physician in the diagnosis of endoscopic diseases, such as gastrointestinal bleeding, using a colour space rotation-based method. This algorithm modifies the images RGB representation channels in such way that the bleeding information can be effectively enhanced in a single channel. Tests in endoscopic images were performed and results have shown that the proposed colour-based rotation approach successfully increases the detection of bleeding regions in endoscopic images, corroborating the usefulness in physician diagnosis assistance.