# A low light video enhancement using interval valued intuitionistic fuzzy set with HVI space

## TL;DR

This paper introduces a new method for improving low-light videos using fuzzy logic and color space transformation, outperforming existing techniques in quality metrics.

## Contribution

A novel Interval-Valued Intuitionistic Fuzzy Generator framework for low-light video enhancement is proposed.

## Key findings

- The IVIFG framework outperforms conventional fuzzy and deep-learning methods in no-reference quality metrics.
- The method preserves illumination, contrast, and perceptual color fidelity effectively.
- Validation on a custom low-light traffic dataset confirms its robustness and practical applicability.

## Abstract

Enhancing low-light videos is essential for applications such as surveillance, autonomous driving, and medical imaging. However, achieving effective contrast improvement remains challenging due to poor illumination, noise, and over-enhancement artifacts. To address these issues, this study presents a novel Interval-Valued Intuitionistic Fuzzy Generator (IVIFG)-based framework for low-light video enhancement. In the proposed approach, the input video is decomposed into individual frames, which are enhanced using the IVIFG model. The enhanced frames are then transformed into the HVI color space, and visually optimal frames are selected using an entropy-based criterion to preserve illumination, contrast, and perceptual color fidelity. To demonstrate practical applicability, the method is applied to a no-reference low-light video and evaluated using standard quality metrics, including entropy, AMBE, CII, NIQE, and BRISQUE. Additional validation is performed on a custom low-light traffic dataset. In both scenarios, the proposed approach is compared with conventional fuzzy methods (IFA, IVIFA+CLAHE, ANV, NIFG) and deep-learning models (Zero-DCE, Zero-DCE++, EFINet, and FlightNet). The results show that the IVIFG framework consistently outperforms existing techniques, achieving superior performance across all no-reference metrics. These findings highlight its robustness and strong potential for real-world deployment in low-light video enhancement applications.