A SYSTEMATIC STUDY AND COMPARATIVE ANALYSIS OF UNDERWATER IMAGE ENHANCEMENT TECHNIQUES WITH EMPHASIS ON FUZZY HISTOGRAM EQUALIZATION

Abstract

A critical issue for marine applications is underwater image enhancement. Light ab sorption and scattering at different wavelengths degrade underwater images producing contrast-loss and color distortion and structural degradation. The Fuzzy Histogram Equalization (FHE) has been considered as a potentially fruitful family of techniques to deal with intensity uncertainties using membershipfunctions. Themajorgapwithinthis area is the absence of a systematic and comparative study on different FHE techniques as well as the lack of clear evidence of what factors actually affects the performance of the FHE algorithms. To fulfill the above purpose, this thesis is two-fold. First, the author conducts a systematic review of twelve FHE-related techniques published between 2017 and 2024, grouping the methods into bi-histogram, weighted, contextual, adaptive, intuitionistic, and type-2 fuzzy approaches. The review reveals that a fixed membership parameter is a consistent central limitation across all the analyzed techniques. Second, based on controlled experiments involving fourteen techniques across four standard datasets, the study concludes that adaptive ability and coupled color correction, not complexity of membership function, determine the performance of FHE. FIMHE automatically sets the clip limit based on the fuzziness of each input image, and thus consistently achieves good structural quality. FCCE achieves higher perceptual quality as it is grounded in context-aware membership functions; however, the loss of structural quality is notably large. In addition, intuitionistic and type-2 fuzzy methods offer no notable structural benefit over standard HE.