BEMD-GI: A BI-DIMENSIONAL EMPIRICAL MODE DECOMPOSITION FRAMEWORK FOR ENHANCING WIRELESS CAPSULE ENDOSCOPY IMAGES

Abstract

Wireless Capsule Endoscopy (WCE) has transformed digestive tract imaging, however technology and environmental factors are degrading image quality.
Traditional ways of improving pictures don't work on WCE images because they aren't static. This article suggests BEMD-GI, a new two-dimensional Empirical Mode Decomposition approach made just for improving WCE images. The framework adaptively breaks down photos into Intrinsic Mode Functions (IMFs) without making any assumptions about the signal properties ahead of time. The results of testing on 2,847 WCE images show big improvements: PSNR went up by 6.2 dB (from 24.63 to 30.83 dB), SSIM went up by 0.145 (from 0.642 to 0.787), and entropy went up by 12.3%. The suggested method works better than five current best methods, even deep learning-based ones. It keeps 15.7% more diagnostic features while still being computationally efficient enough for clinical processes.