Most image enhancement methods are developed based on a tacit assumption: noise is an undesirable artifact to be suppressed before or during enhancing. Be that as it may, this assumption does not hold for real-world noise, which has a structural characteristic of its own. It's not constant in space, nor does it present a spectrally random pattern; instead, it reflects uncertainties about local situations and signal reliability in collected data. Typical methods which treat noise as an interference instead of information often bring out over-enhancement or fake textures, missing original detail. In this paper, we provide a framework of noise-aware image enhancement that interprets signal disturbance as structural evidence rather than being an anomaly. The process then becomes redefined from intensification to making decisions on the basis of noise reliability. To implement this, a combined spatial-frequency domain analysis is employed to study noise behaviour both locally and globally. By fusing these complementary noise markers, we give selective enhancement: structures which are firmly based will be built up while enhancement is held back in zones heavily dominated by noise. In further development, this paper introduces a hybrid analytical-learning extension. A lightweight neural module learns the degree to which enhancement should be applied, while being explicitly conditioned on noise descriptors that are understandable. Experimental results show that the method proposed in this paper achieves stable and visually comprehensive enhancement under conditions of challenging noise: it maintains a high degree of fidelity to structure and does not add noise. This work will lead to a conceptual change in future enhancement design: constituted as an essential principle rather than some way of eliminating the only concerned factor.
25 January 2026 in International Journal of Computers and Informatics
In such environment fraught with both conflict and scarcity of resources, images captured by surveillance cameras, mobile phones, digital cameras or unmanned aerial vehicles (UAV) tend to encounter multiple degradation types such as noise, blur, deficient illumination, and artifact in transmission. Typical image enhancement techniques and supervised deep learning models require large paired (low/high-quality) datasets. However, these are rarely obtainable under the restricted conditions of post-conflict environments. This paper proposes a new damage-aware, self-supervised image enhancement framework based on the architecture of Vision Transformer (ViT). The proposed model implicitly learns different forms of degradation and autonomously enhances image quality without the need for high-resolution reference images as ground-truth. A multi-task self-supervised objective that allows simultaneous recognition and restoration of degradation. Experimental results show that our technique improves perceptual quality, structural integrity, and robustness by a large margin, when contrasted with conventional CNN-based methods. Of significance is the fact that the algorithm is highly effective in its ability to handle unseen real-world and post-conflict degradation scenarios.