Framework for the Restoration of Capsule Endoscopy Images Using Partial Differential Equations-based Filter

Abstract

This study presents a design and experimental results of a novel noise reduction framework for capsule endoscopy (CE) images. CE images exhibit random noise, motion blur, and illumination related degradations. These degradations reduce the efficiency of the computer-aided diagnosis (CAD) system. Restoration process aims to minimize degradations as much as possible; however; measurement of the degradations is a challenge. This study proposes a partial differential equation-based filter for the restoration of CE images. The proposed system performs noise removal while retaining high-frequency components. It measures the performance of the proposed restoration framework based on metrics related to image quality. Visual and statistical results show significant improvement in CE image quality. The proposed approach is capable of noise removal while preserving a low level and a few high-level components leading to better results. Finally, it also presents the performance of a CAD system with and without restoration. © 2023 IETE.