Bibliographic record
Abstract
Image denoising is an inseparable pre-processing step of many image processing algorithms. Two mostly used image denoising algorithms are Nonlocal Means (NLM) and Block Matching and 3D Transform Domain Collaborative Filtering (BM3D). While BM3D outperforms NLM on variety of natural images, NLM is usually preferred when the algorithm complexity is an issue. In this thesis, we suggest modified version of these two methods that improve the performance of the original approaches. The conventional NLM uses weighted version of all patches in a search neighbourhood to denoise the center patch. However, it can include some dissimilar patches. Our first contribution, denoted by Similarity Validation Based Nonlocal Means (NLM-SVB), eliminates some of those unnecessary dissimilar patches in order to improve the performance of the algorithm. We propose a hard thresholding pre-processing step based on the exact distribution of distances of similar patches. Consequently, our method eliminates about 60% of dissimilar patches and improves NLM in terms of Peak Signal to Noise Ratio (PSNR) and Stracuteral Similarity Index Measure (SSIM). Our second contribution, denoted by Probabilistic Weighting BM3D (PW-BM3D), is the result of our thorough study of BM3D. BM3D consists of two main steps. One is finding a basic estimate of the noiseless image by hard thresholding coefficients. The second one is using this estimate to perform wiener filtering. In both steps the weighting scheme in the aggregation process plays an important role. The current weighting process depends on the variance of retrieved coefficients after denoising which results in a biased weighting. In PW-BM3D, we propose a novel probabilistic weighting scheme which is a function of the probability of similarity of noiseless patches in each 3D group. The results show improvement over BM3D in terms of PSNR for an average of about 0.2dB.
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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".