Performance Evaluation of various Image Enhancement Techniques for Gastric Polyp Detection in the Kvasir Dataset

Unusual growths found in the lining of the stomach are referred to as gastric polyps, the presence of such can be detected using upper endoscopic imaging. Detecting stomach polyps early is very crucial to prevent gastrointestinal cancers. Endoscopic imaging is a standard diagnostic tool recommended by medical practitioners to investigate the upper intestinal tract. This procedure often suffers from inconsistencies in illumination, color balance and contrast, hindering polyp identification. The current work investigates the impact of various image preprocessing and image enhancement techniques on the Kvasir dataset, a publicly available collection of gastric endoscopic images. This work evaluates techniques like resizing, rescaling, normalization, contrast enhancement, color space conversion (Lab), contrast stretching, Contrast Limited Adaptive Histogram Equalization, unsharp masking, and Histogram equalization. The effectiveness of these techniques is assessed through qualitative and quantitative analysis and their impact on polyp visibility for improved computer-aided diagnosis systems. Quantitative analysis based on evaluation metrics like Peak Signal-To-Noise Ratio, Mean Squared Error and Structural Similarity Index, is calculated and a comparative study for the various enhancement techniques was done in this work.