MVG-VQA
No-reference video quality assessment fusing spatial MVG features with temporal features via XGBoost regression
A no-reference VQA system fusing spatial MVG features (Log-Gabor, LBP, edge statistics) with temporal features (frame-to-frame motion, flicker) via XGBoost regression. Outperforms standard metrics on compression artifact and noise benchmarks across multiple NR-VQA datasets. Features were extracted in Python with hyperparameters tuned via cross-validation.