Sequential Low-Light Image Enhancement Using Eigenvalue-Based Gamma Correction and Laplacian Feature Map-Based Transmission Map

Low-light images are typically captured under adverse conditions, such as nighttime environments or low-contrast settings, which severely hinder visual understanding and object recognition. Although various deep learning and image signal processing techniques have been proposed to address this issue, low-light image enhancement remains a challenging task due to the lack of ground-truth reference images and the diverse range of real-world degradation phenomena. To overcome these limitations, this paper proposes a novel low-light image enhancement framework utilizing an eigenvalue-based gamma correction method based on the ratio of the 1st and 2nd eigenvalues, which effectively captures the inherent characteristics of the input image. Furthermore, because image enhancement processes inevitably amplify background noise, this paper incorporates a specialized denoising method inspired by dehazing theory to mitigate this side effect. A critical challenge in mathematical dehazing models is the accurate estimation of the transmission map, which is often distorted by strong light components in low-light scenes, leading to undesirable halo effects and color distortion. To solve this, this paper introduces a Laplacian feature map-based transmission map, leveraging the fact that Laplacian edge components remain independent of ambient illumination. Extensive experimental evaluations demonstrate that the proposed method outperforms state-of-the-art approaches, delivering superior visual quality in both objective and subjective assessments.