Joint Self-Supervised Enhancement and Denoising of Low-Light Images
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Images taken under low-light conditions often suffer from multiple degradations such as low visibility and unknown noise. Low-light image enhancement is an important task in the field of computer vision. In order to avoid the limited number of samples in paired datasets, several self-supervised enhancement methods have been developed. However, due to the designed illumination gradient prior, most self-supervised enhancement methods based on Retinex cannot effectively constrain the illumination or suppress the amplified real noise. To solve this problem, this paper explores a joint self-supervised enhancement and denoising method for low-light image. Initially, we proposed a new regularization term, named TV-Huber, and developed an adaptive illumination estimation network (AIE-Net) to explore the intrinsic relationship between structure and texture in the illumination map. Next, the camera response model and the learned illumination are then used to enhance the contrast of low-light images and mitigate color shifts. Finally, the learned illumination maps are transformed into illumination masks. Under the assumption of independent and zero-mean noise, selective feature injection is performed on the shallow features extracted by the blind-spot network (BSN) to reduce information loss while removing unknown real noise in the dark area. Extensive experiments show that the proposed method has good generalization ability on five challenging low-light image datasets and outperforms other methods in terms of visual quality and quantitative comparison.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it