Abstarct: Wireless Capsule Endoscopy (WCE) is a non-invasive medical imaging device that doctors utilize to examine the digestive system. Most WCE images have Specular Reflection (SR). The SR is frequently an effect of the rigorous light and luminous regions appearing on WCE frxames, influencing the performance of abnormality detection methods and physician analysis. The current investigation aims to develop an approach that can automatically segment and eliminate the SR of WCE images to improve the accuracy of abnormality detection approaches and help doctors to make a powerful diagnosis. In this study, we present a novel method for SR elimination by the tiniest damage to the image texture using the Homomorphic Filter (HF). The proposed method contains three steps. Initially, we use a strong segmentation technique called the U-Net model to segment SR baxsed on semantic segmentation. Afterwards, we use the HF to separate the illumination and reflection components. Damage stemming from SR includes various effects on these two components. Our proposed method enriches each of these components separately instead of enhancing the whole image uniformly. Finally, we reconstruct the SR areas using Navier-Stokes (NS) inpainting technique baxsed on fluid dynamics. Our experiments were performed on three various types of WCE datasets. The evaluations were performed with the following criteria accuracy, false negative rate, false positive rate, precision, recall, and F-Measure. The results reveal that our proposed method quantitatively improved the accuracy of abnormality detection methods and improved their performance, and qualitatively eliminated the SR admirably.