Enhancement Algorithms to Improve the Class of Endoscopy Images to Advance Gastrointestinal Track Disease Detection

Abstract

Gastro Intestinal (GI) track diseases are major health concern which require accurate detection for early diagnosis and treatment. One of the primary methods of GI track disease detection is to use endoscopic images of digestive track system. However, these images often are degraded by the presence of noise and poor contrast, which directly affect the diagnostic accuracy. Enhancing the quality of the degraded endoscopic images can greatly help physicians during diagnostics and treatment plan. In this work a unified approach which combines denoising and contrast variation correction is proposed. Denosing is executed using a fusion algorithm that pools discrete wavelet transformation and singular value decomposition that is combined with non-local means denoising algorithm. The contrast variation correction is performed using an adaptive gamma correction algorithm enhanced using particle swarm optimization method. Experiments proved that the proposed unified approach improves the visual quality while preserving significant image, edge and structure details. This method can therefore be used safely by GI track disease detection and classification system to improve its diagnostic accuracy.