Weighting and thresholding-based detail preserving image enhancement of gastrointestinal images with noise suppression

Abstract

Endoscopy is a widely practiced technique across the globe, utilized to envision the gastrointestinal (GI) tract of the human body without causing harm. In GI endoscopy, images affected by noise may not be helpful to support a visual diagnosis of the mucosal surface at good quality. In such advanced imaging systems, image quality enhancement becomes particularly beneficial. In this study, a non-invasive numerical method of image enhancement is examined to improve the image quality metrics as is critical in GI imaging. A unique approach of weighting and thresholding is examined in this study, where the weighting and thresholding play a significant role in image enhancement. Deviations from the proposed weighting scheme results in the reduction of the peak signal to noise ratio (PSNR) among other key metrics. The proposed method evaluated on eight datasets are expounded to produce optimal performance, particularly in that of the PSNR, entropy and contrast over the state-of-the-art (SOA) enhancement algorithms. The proposed method yields a universal image quality index of 93 %, a structural similarity of 86 % and that of 25 %, 11 %, 19 % and 30 % of PSNR, entropy, contrast, and distortion metric enhancements on average in comparison with the SOA, contributing to the improved quality of the resultant frame aiding to view image features at higher detail and overall quality, both quantitively and qualitatively. This in conjunction with higher optical magnification and faster graphics processing units (GPU’s) may be beneficial for a real-time examination of the mucosal surface of the GI tract at superior quality.